Multi-Modal Clinical Session Recording System
Platform
RK3588 / Linux
Type
Edge Video / Embedded
Encoding
H.264 HW Accel
Status
Delivered
Background
Synchronized Multi-Modal Edge Recording — Visible + Thermal Video, Hardware-Accelerated, on a Single Embedded Platform
This project centered on building a reliable, multi-source recording pipeline on an RK3588-based embedded Linux platform. The system captures synchronized video from two CSI visible cameras and a USB thermal camera simultaneously, compresses footage using hardware-accelerated H.264 encoding, organizes everything into structured session folders via physical button control, and automatically uploads completed recordings to cloud storage. The result is a fully autonomous, operator-friendly capture system that requires no keyboard or technical user to operate — purpose-built for clinical session recording, industrial inspection, and edge video applications where precision, reliability, and offline-first operation are non-negotiable.
Challenges
Key Project Challenges
1
Multi-Source Capture Reliability
Simultaneously driving two CSI visible camera streams and a USB thermal camera on a single embedded platform without stream conflicts, dropped frames, or pipeline stalls at 30 FPS.
2
Hardware-Accelerated H.264 Encoding on RK3588
Integrating the RK3588's hardware video encoder into the Linux camera pipeline to compress visible streams in real time without burdening the CPU — critical for sustained, lossless multi-channel recording.
3
Session Data Organization Without Operator Input
Designing a button-triggered recording state machine that correctly creates, names, and closes session directories on each start/stop cycle — ensuring zero file mix-ups across consecutive recording sessions.
4
Network-Independent Cloud Archival
Building a cloud upload pipeline that runs after local recording completes so capture is never blocked or disrupted by network availability — ensuring every session is archived without compromising recording integrity.
Solutions
How We Built It
Our Approach
Full Edge Pipeline — Dual CSI + USB Thermal Capture, HW H.264 Encoding, State-Machine Recording, and Automated Cloud Archival on RK3588 Linux
The system was architected as a tightly integrated embedded Linux pipeline with clearly separated capture paths and a central recording state machine. Both CSI visible cameras initialize and stream independently, with their output routed through the RK3588's hardware H.264 encoder to keep CPU load minimal and encoding latency predictable. The USB thermal camera is captured through a dedicated USB path and stored in parallel. A physical button interface drives a recording state machine that starts the pipeline, creates a uniquely named session directory, writes all streams into it, and cleanly closes everything on stop. Once a session is finalized locally, an automated cloud upload service archives the completed folder — decoupling network dependency entirely from the recording process. The full pipeline was validated across camera bring-up, synchronized capture, encoding output, physical control behavior, session folder integrity, and upload completion.
Benefits
Value Delivered
Synchronized Multi-Modal Capture
Visible and thermal streams are captured simultaneously and stored in the same session directory, giving every recording a complete, time-aligned multi-modal evidence package.
Hardware-Accelerated Encoding
H.264 encoding runs on the RK3588's dedicated hardware block, keeping CPU usage low and enabling sustained 30 FPS visible capture across both cameras without thermal throttling.
Operator-Free Physical Controls
A physical button starts and stops the entire recording workflow — no keyboard, no screen, no technical user required — making the system deployable in clinical and field environments with zero training overhead.
Zero-Mix Session Organization
Each recording cycle automatically creates a uniquely named session folder, eliminating any risk of file collisions or data mix-ups across consecutive sessions — critical in clinical and compliance-sensitive use cases.
Offline-First, Cloud-Ready Archival
Recording never depends on network availability. Sessions are captured and stored locally first, then uploaded to cloud storage automatically once complete — ensuring no footage is lost due to connectivity issues.
Reusable Architecture for Multiple Verticals
The same pipeline applies directly to industrial inspection, machine monitoring, safety systems, training capture, and any edge product requiring organized, uploadable multi-camera recordings.
Client Feedback
What the Client Said
"
The hardest part was getting all three cameras to record at the same time without anything dropping out. That turned out to be a non-issue — the pipeline handled it cleanly from the start. The button control was exactly what we needed; our operators don't touch a keyboard during sessions. What really impressed us was how the session folders came out perfectly organized every single time, and recordings just appeared in cloud storage without us doing anything. For a clinical environment, that level of reliability is everything.
Need a Similar System?
We build custom embedded Linux, edge video, and camera integration systems tailored to your requirements.
Request a Free Quote →