Revasri Thermal Vision (Android)

An open-source Android application and USB driver for the REVASRI R-T160, NOHAWK N-T160, and ThermPal USB-C thermal imaging cameras (0x04b4:0x000a).

---

Why This Project Exists

Budget USB-C thermal cameras like the REVASRI R-T160 (and OEM variants from Nohawk and Inreii) offer outstanding sensor performance for their price, but their official companion app (Camera+ / com.inreii.neutralapp):

1. Is not readily available on Google Play.
2. Bundles legacy 32-bit (armeabi-v7a) native .so libraries that fail to run on modern 64-bit-only phones (Pixel 7/8/9/10, Galaxy S23/S24/S25).
3. Lacks 16 KB ELF segment alignment, triggering immediate OS memory-fault aborts on Android 15, 16, and 17 beta.
4. Standard UVC (USB Video Class) camera viewers cannot stream from this device because it communicates over a vendor-specific bulk endpoint protocol.

This project provides a 100% pure Kotlin implementation using Android's native android.hardware.usb API:

• Zero NDK or C/C++ libraries (.so): Immune to 32/64-bit ABI conflicts and 16 KB page-size enforcement.
• Runs natively on Android 15, 16, and 17 beta.
• Modern, responsive Jetpack Compose interface with hardware-accelerated rendering.

---

Hardware & Protocol Specifications

Through protocol reverse-engineering of USB descriptors and bus packets:

Attribute	Specification
USB Identifiers	Vendor ID: 0x04b4 (Cypress), Product ID: 0x000a
USB Class	Class 255 (Vendor Specific), Subclass 240, Protocol 1 (com.white.ThermPalProtocol)
Interface & AltSetting	Interface 0, Alternate Setting 1 (AltSetting 0 contains 0 endpoints)
Bulk IN Endpoint	0x82 (Packet size 512 bytes)
Bulk OUT Endpoint	0x06 (Packet size 512 bytes)
Stream Frame Rate	~24.0 – 25.6 FPS (~41.6 ms period)
Readout Matrix	256 columns × 196 rows (192 sensor rows + 4 telemetry rows)
Active Sensor Area	160 × 120 pixels (Centered at Rows 36..155, Columns 48..207)

Control Commands (Bulk OUT to 0x06)

• Init Handshake: 0x80 0x81 (Sent once upon connection to start bulk streaming)
• Mechanical Shutter Trigger: 0x80 0x00 (Actuates internal solenoid for NUC dark-frame calibration)
• High Gain Mode (-20°C to 150°C): 0x80 0x21
• Low Gain Mode (Up to 450°C): 0x80 0x20

Framing Protocol

Each video frame is transmitted across 7 sequential bulk packets of 14,348 bytes each:

• 12-byte header: Contains magic (0x0c8c/0x0c8d), timestamp, and line offset (1, 29, 57, 85, 113, 141, 169).
• 14,336-byte payload: 28 lines × 256 pixels of 16-bit little-endian raw bolometer ADU readings.
• Telemetry (Rows 192..195): Serial number, model identifier (R-T160), firmware version, and real-time die/shutter temperature sensors (Row 193).

---

Calibration & Radiometry Engine

Verified Two-Phase Shutter Calibration

Unlike standard consumer apps that merely click the shutter mechanically without synchronizing software, Revasri Thermal Vision executes a closed-loop calibration pipeline:

1. Solenoid Transit Handling: When Calibrate is tapped, the hardware shutter command is dispatched, and transient transit frames (~80 ms) are automatically bypassed.
2. Phase 1: Stabilization & Temperature Calibration:
   • Observes consecutive frame mean delta (\le 12 ADU).
   • Verifies stability across 3 consecutive readings with the shutter paddle fully seated.
   • Anchors the radiometric baseline reference to the internal unit die temperature (T_{\text{unit}} = \text{fpaTemperatureC}).
3. Phase 2: Video Calibration:
   • Accumulates closed-shutter frames to compute a 2D spatial Fixed-Pattern Noise (FPN) matrix.
   • Eliminates microbolometer vertical stripes, column offsets, and static grain.
4. Verification & Persistence:
   • Verifies consistency between video mean and temperature baseline (|\Delta| < 150 ADU), verifies bolometer noise spread, and validates temperature sanity before committing.
   • Saves verified baseline to thermal_fpn_baseline.bin across app launches.

Radiometric Temperature Calculation

Pixel temperature is calculated via linear microbolometer radiometry:

T_i = baselineTempC + (pixelAdu_i - baselineMeanAdu) / (Sensitivity * Emissivity) + userOffsetC

• Default Offset Constant: Includes a defined constant DEFAULT_CAMERA_TEMP_OFFSET_FAHRENHEIT = +22.0°F (+12.22°C) to align baseline radiometry with ambient environmental readings.
• Material Emissivity: Configurable from 0.10 to 1.00 (defaults to 0.95 for skin, wood, drywall, and matte surfaces).

---

App Features

• Fluid Live Thermal Stream: 24+ FPS real-time rendering via Jetpack Compose.
• Spot Tracking HUD:
  • Red crosshair on hottest point with peak temperature readout badge.
  • Cyan crosshair on coldest point with minimum temperature readout badge.
  • Green crosshair on center point with live temperature badge.
• Orientation: Defaults to 90° orientation; one-tap rotation cycling (0°, 90°, 180°, 270°).
• Temperature Units: Automatic locale detection (Celsius / Fahrenheit) with one-tap toggle.
• Color Palettes:
  • Inferno
  • Ironbow
  • Rainbow
  • White Hot (Grayscale)
  • Black Hot (Inverted Grayscale)
• 1-Tap Calibrate & Options:
  • Tapping Calibrate immediately fires the verified shutter calibration.
  • Dropdown arrow provides quick access to Fine-Tune Temperature, Guide, and Calibration Reset.
• Interactive Guide: Built-in explainer dialog (?) covering calibration, emissivity, and thermal tips.
• Screen Wake-Lock: Automatically holds screen awake (FLAG_KEEP_SCREEN_ON) while the camera is active.
• Persistent Preferences: Automatically saves palette, rotation, gain mode, temperature units, emissivity, and calibration to SharedPreferences.
• Snapshot Capture: Exports uncompressed PNG captures directly to device gallery (Pictures/RevasriThermal).

---

Building & Installing

Prerequisites

• JDK 17+
• Android SDK 35 (Android 15)

Build Debug APK

./gradlew assembleDebug

Output APK: app/build/outputs/apk/debug/app-debug.apk

Install via ADB

adb install -r app/build/outputs/apk/debug/app-debug.apk

Launch App

adb shell am start -n com.adamoutler.revasri/.MainActivity

---

License

Developed by Adam Outler (adamoutler@gmail.com).

Licensed under the Apache License, Version 2.0. See LICENSE for details.