Spatial Computing Experiments
Working through multi-camera depth sensing, point clouds, and real-time spatial analysis.
Spatial Computing Experiments
Related project: Multi-Camera Vision & Point Cloud Experiments
This started as a fairly simple idea and turned into a much deeper rabbit hole once synchronization, calibration, and USB bandwidth became real constraints.
Most of the work here revolves around trying to combine multiple depth cameras into a usable real-time spatial system.
Current areas of experimentation include:
- multi-camera calibration
- point cloud alignment
- depth filtering
- planar detection
- object boundary detection
- visualization pipelines
- sensor synchronization
- GPU processing experiments
One thing that became obvious pretty quickly was that getting clean data from multiple cameras mattered a lot more than the actual measurement logic.
A large amount of effort ended up going into:
- camera placement
- USB topology
- timing consistency
- transform alignment
- visualization tools
- debugging noisy depth data
The current setup uses several Intel RealSense cameras in different arrangements depending on what is being tested.
Some experiments involve opposing angled views while others focus on top-down reference geometry.
A lot of the work is still highly experimental and changes regularly.
Future directions probably include:
- distributed processing
- AI-assisted segmentation
- improved calibration tooling
- web-based visualization
- better GPU acceleration
- automated scene reconstruction
The project continues to evolve mostly as a general spatial computing and sensor fusion sandbox.