DEV Community

Pavan Kumar Sadashiv
Pavan Kumar Sadashiv

Posted on

MATSYA-Node: Complete System Architecture, Hardware Selection & Working Mechanism

#iot #edgeai #hardware #marine

MATSYA-Node: Master Hardware & Operational Specification

Summary for the Team: This document brings together our entire technical solution for the AquaAirX MATSYA Challenge 2026 into a single, cohesive engineering blueprint. It explains every device we use, the exact physical working mechanism, why each component was chosen over alternatives, and how it handles real ocean operations.

1. High-Level System Architecture

MATSYA-Node is a portable, zero-boat-modification marine sensing and Edge AI logging system. It operates completely offline at sea and syncs automatically when returning to coastal connectivity.

==================================== SEAWATER ====================================
 [UNIT 1: WET SENSOR POD]                       [UNIT 2: NET DEPTH CLIP]
 • Maxim DS18B20 Temp Probe                     • MS5837-30BA Pressure & Depth
 • Optical Turbidity Probe (TS-300B)            • Deep Thermocline Sensor
 • Piezo Hydrophone                             • Saltwater Immersion Pins
 • Saltwater Immersion Pins                     (Rides with Net down to 300m)
 (Rides 0.5m-1.0m below surface)                        │
          │                                             │
          │ (Polyurethane Kevlar Umbilical Cable)       │ (Tethered / BLE Log)
          ▼                                             ▼
=================================== BOAT DECK ====================================
                       [TIER 2: IP67 DECK EDGE BOX]
 ┌───────────────────────────────────────────────────────────────────────────────┐
 │ • Brain: Raspberry Pi Zero 2W / ESP32-S3 Dual-Core NPU                        │
 │ • Vision: Sony IMX219 8MP Wide-Angle Camera (Overhead Fish Sorting Deck)      │
 │ • Navigation: u-blox NEO-M8N GPS Module                                       │
 │ • Kinematics: MPU6050 6-DOF IMU (Drift, Roll, Engine Vibration Classifier)     │
 │ • Radio Mesh: SX1262 LoRa Transceiver (865-867 MHz ISM Band)                  │
 │ • Power: 12V Boat Battery Tap + 10W Solar Trickle + 10,000mAh LiFePO4 Battery │
 └──────────────────────────────────────┬────────────────────────────────────────┘
                                        │
                         (Local Wi-Fi AP / Bluetooth Low Energy)
                                        │
                                        ▼
                   [TIER 3: FISHERMAN VERNACULAR COMPANION]
 ┌───────────────────────────────────────────────────────────────────────────────┐
 │ • Pre-cached INCOIS Potential Fishing Zone (PFZ) Satellite Maps (Offline)     │
 │ • Real-time Vernacular Voice Guidance (Tulu, Kannada, Malayalam, Tamil)      │
 │ • Automated Trip Log & Crew Settlement Ledger                                │
 │ • Dual-Ledger Privacy Vault (Exact GPS encrypted locally; never exposed)      │
 └──────────────────────────────────────┬────────────────────────────────────────┘
                                        │
                  (Store-and-Forward sync via 4G / Harbour Wi-Fi on return)
                                        │
                                        ▼
================================ HARBOUR CLOUD ===================================
                 [TIER 4: SCIENTIFIC INTELLIGENCE (CMFRI / INCOIS)]
 ┌───────────────────────────────────────────────────────────────────────────────┐
 │ • Uber H3 Hexagonal Spatial Indexing (Res 5: ~8-10 km grid anonymization)     │
 │ • PostGIS & TimescaleDB Spatiotemporal Database                              │
 │ • Catch Per Unit Effort (CPUE), Biomass Density & Juvenile Bycatch Analytics  │
 └───────────────────────────────────────────────────────────────────────────────┘
Enter fullscreen mode Exit fullscreen mode

2. Complete Hardware Bill of Materials: Device Selection & Rationale

Here is the exact hardware list, explaining what device is used, why it was chosen over alternative components, and its estimated cost:

Subsystem Exact Component Chosen Alternative Considered Why This Component Was Chosen Unit Cost (INR)
Edge Compute Raspberry Pi Zero 2W (Quad-core 64-bit ARM @ 1.0 GHz, 512MB RAM) or ESP32-S3 Raspberry Pi 4B / Jetson Nano Pi 4 and Jetson draw 10W–15W of power, overheat in sealed boxes, and cost ₹6,000–₹15,000. Pi Zero 2W consumes only 0.7W–1.2W, boots quickly, runs TensorFlow Lite / INT8 YOLO models, and costs under ₹1,700. ₹1,650
Deck Vision Sony IMX219 8MP Sensor with 120° Wide-Angle Lens & Acrylic Dome USB Webcams / Standard Pi Camera v1 (OmniVision OV5647) USB webcams have loose cables that shake loose and high USB bus latency. OV5647 has poor low-light response. IMX219 offers excellent dynamic range for bright sunlight on white deck fiberglass and low-light sorting at dawn/dusk. ₹650
Depth & Net Trigger TE Connectivity MS5837-30BA (I2C Pressure Sensor) Mechanical float switch / Ultrasonic depth gauge Mechanical floats jam from sand and fish scales. Ultrasonic sensors fail at sea surface bubbles. The MS5837 is hermetically gel-filled, immune to saltwater corrosion, rated to 30 bar (300m depth), and detects submersion down to 2 mm accuracy. ₹550
Water Temperature Maxim DS18B20 (Encapsulated in 316 Stainless Steel Probe) Analog NTC Thermistor / PT100 RTD Thermistors require analog op-amp circuits that drift with boat voltage fluctuations. DS18B20 communicates via a digital 1-Wire protocol, allowing 10-meter cable runs without signal degradation or noise pick-up. Grade 316 SS resists pitting corrosion. ₹180
Water Clarity / Plankton Optical Turbidity Sensor Probe (TS-300B Infrared 940nm) Secchi disk manual estimation / Commercial fluorometer Secchi disks require human effort. Commercial fluorometers cost over ₹80,000. The TS-300B uses infrared light attenuation to measure Nephelometric Turbidity Units (NTU) continuously, highlighting upwelling zones. ₹220
Passive Acoustics Piezoelectric Contact Hydrophone (Lead Zirconate Titanate potted in Polyurethane) Air microphone / Hydrophone array Air mics pick up wind howl. Commercial hydrophone hydro-arrays cost ₹1,00,000+. A simple polyurethane-potted piezo element listens to underwater sound frequencies (20 Hz - 40 kHz) to record engine noise, fish acoustics, and echo pulses. ₹200
Navigation u-blox NEO-M8N GPS Module with Active Ceramic Patch Antenna Generic NEO-6M NEO-6M only tracks US GPS and takes up to 4 minutes for a cold fix. The NEO-M8N concurrently tracks GPS, GLONASS, Galileo, and BeiDou, securing satellite locks in under 26 seconds even with boat roll and mast occlusions. ₹480
Boat Kinematics InvenSense MPU6050 (3-axis Gyroscope + 3-axis Accelerometer) Separate inclinometer / Engine tachometer wiring Tapping into boat engine tachometers voids warranties and differs across boat models. MPU6050 measures vessel acceleration, roll, pitch, and deck vibration directly, allowing software to determine boat state (Steaming vs. Trawling vs. Idle) non-invasively. ₹150
Offshore Mesh Radio Semtech SX1262 LoRa Transceiver (865–867 MHz ISM Band) Satellite Transceivers (Iridium/Inmarsat) / Cellular 4G Satellite data subscriptions cost ₹3,000–₹8,000 per month per boat (unaffordable for small fishermen). Cellular 4G drops 10–15 km from shore. SX1262 provides 5–15 km license-free boat-to-boat mesh communications with zero monthly bills. ₹420
System Power 12V-to-5V Buck Converter + 10,000mAh LiFePO4 Cell + 10W Solar Panel Standard Li-ion 18650 power bank Standard Li-ion cells degrade in high engine-room heat (45°C–50°C) and can thermal runway. LiFePO4 chemistry is chemically stable up to 60°C, lasts 2,000+ cycles, and provides 18+ hours of uninterrupted buffer power. ₹600
Enclosure & Mount IP67 Polycarbonate Box + 316 SS C-Clamp + Cable Glands 3D printed PLA box / Zip ties PLA warps under tropical sun and absorbs moisture. Our enclosure uses industrial UV-stabilized polycarbonate, EPDM gaskets, PG7 marine glands, and heavy-duty 316 stainless-steel clamps for rapid 60-second tool-less boat mounting. ₹350
TOTAL BOM Complete Turnkey MATSYA-Node — Affordable, scalable to 10,000+ boats ₹5,450 (Drops to ₹3,600 @ scale)

3. How the Physical Hardware Actually Operates in the Field

Step 1: Mounting in 60 Seconds (Zero Hull Alterations)

  1. The Deck Edge Box: Clamped to the rear wheelhouse cabin roof overhang, canopy pipe frame, or winch stanchion using a quick-tightening 316 stainless-steel C-clamp. The box is elevated 1.8m to 2.5m above the deck, angled downward at 45° to 60° to view the fish collection/sorting area.
  2. The Wet Sensor Pod: Connected to the deck box via a durable polyurethane umbilical cable and suspended over the side (gunwale) or clamped to the stern transom bracket, submerged 0.5m to 1.0m into seawater.
  3. The Net Depth Clip: Clipped with a stainless-steel carabiner to the headrope or wing of the fishing net.

Step 2: Automatic Net-Drop & Haul Detection (Zero Buttons Needed)

Fishermen will not press buttons while handling gear. The system detects fishing activity using Triple-Sensor Fusion:

  1. Submersion Detection: When the net is thrown into the sea, the MS5837-30BA pressure sensor registers an increase from 1.0 bar (air) to $>1.05 ext{ bar}$ ($>0.5 ext{m}$ depth). Gold-plated conductivity pins confirm saltwater contact.
  2. Kinematic Confirmation: The NEO-M8N GPS registers vessel speed dropping from cruising speed (7–10 knots) to shooting/towing speed (1.5–3.5 knots), while the MPU6050 IMU detects the characteristic hull roll and high engine torque vibration of a towed net.
  3. Autonomous Logging: The system automatically logs EVENT: NET_DEPLOYED, records the timestamp, GPS location, surface water temperature, and begins tracking haul duration and depth trajectory.
  4. Haul Retrieval: When the net is hauled onto the deck (pressure drops back to 1.0 bar and winch vibrations stop), the system logs EVENT: NET_RETRIEVED and sends a wake-up signal to power up the overhead deck camera.

Step 3: Determining Catch Species & Weight Without Scales

Traditional mechanical or electronic scales corrode in seawater, lose calibration as the boat pitches in waves, and slow down busy deck crews. Instead, MATSYA-Node uses a Dual-Mode Optical Biomass Pipeline:

Mode A: Individual Fish (Length-Weight Biological Relationship)

  1. Focal Calibration: Because the camera is mounted at a fixed distance and angle, every pixel corresponds to a known millimeter dimension (e.g., $1 ext{ px} = 0.8 ext{ mm}$).
  2. Computer Vision Segmentation: As fish land on the sorting deck, an INT8-quantized YOLOv8n-seg model running on the Raspberry Pi identifies each fish, outlines its contour, and extracts its fork length ($L$ in cm) from mouth to tail.
  3. Biological Formula ($W = a \cdot L^b$): Marine fisheries biology (standardized by ICAR-CMFRI) proves that fish weight follows an allometric growth equation. The system calculates weight using species constants for Indian waters:
    • Indian Mackerel (Rastrelliger kanagurta): $W = 0.0051 imes L^{3.12}$
    • Oil Sardine (Sardinella longiceps): $W = 0.0063 imes L^{3.04}$
    • Seer Fish (Scomberomorus commerson): $W = 0.0078 imes L^{2.92}$
    • White Prawn (Fenneropenaeus indicus): $W = 0.0082 imes L^{2.98}$

Mode B: Bulk Catch & Crates (Volumetric Packing Density)

  1. For schooling pelagic catches where hundreds of small fish are loaded into standard 35–40 kg plastic crates (~50–60 L volume), individual counting is unnecessary.
  2. The vision system segments the crate perimeter, measures fill level percentage, and computes the occupied volume ($V$ in litres).
  3. Weight is calculated using bulk packing density ($ ho$): $$ ext{Weight (kg)} = V imes ho_{ ext{species}}$$ (Fresh wet sardines and mackerel pack at $0.85 ext{ to }0.90 ext{ kg per litre}$ of crate volume).

Step 4: Delivering Immediate Fisherman Value (Why They Use It)

  • 15–25% Fuel Savings: Before leaving the harbour, the companion mobile app caches the latest INCOIS Potential Fishing Zone (PFZ) satellite maps. Out at sea (with zero internet), the app fuses this map with real-time temperature from the DS18B20 sensor. When the boat crosses a productive thermal gradient, a vernacular voice alert (in Tulu or Kannada) notifies the skipper: "Target thermal front reached. Fish density high here."
  • Free Boat-to-Boat Emergency Mesh: The SX1262 LoRa module maintains a peer-to-peer radio link with other boats within 5–15 km. If a vessel takes on water, suffers an engine breakdown, or spots sudden squalls, an SOS alert propagates across the boat mesh without expensive satellite airtime.
  • Instant Digital Trip Sheet: Once sorting ends, the app generates a verifiable summary showing fuel burned vs. total species weight caught. This allows the skipper to divide revenue among crew members transparently and provides digital proof of income for formal bank loans and boat insurance.

Step 5: Dual-Ledger Privacy Architecture (Why They Trust It)

Fishermen fiercely protect their secret fishing spots. MATSYA-Node handles data with a strict architectural boundary:

[ RAW SENSOR & CATCH DATA ON EDGE BOX ]
                    │
       ┌────────────┴────────────┐
       ▼                         ▼
[ PRIVATE LOCAL VAULT ]   [ SCIENTIFIC CLOUD LEDGER ]
 • Exact GPS (±2 meters)   • Anonymized Uber H3 Hexagon (Res 5: ~8-10 km)
 • Exact net coordinates   • Aggregated catch volume & species ratio
 • Stays on fisherman's    • 6-Hour time-window bucketing
   phone under AES-256     • Synced to CMFRI / INCOIS upon harbour return
 • NEVER leaves vessel     • Zero competitor can locate secret reefs!
Enter fullscreen mode Exit fullscreen mode

4. How MATSYA-Node Fulfills the 5 Competition Ocean Pillars

Competition Requirement Hardware / Software Module Tangible Output Delivered
1. WHAT is present? Sony IMX219 Camera + Edge YOLOv8-seg + Hydrophone Species composition, biodiversity index, and juvenile bycatch ratio
2. WHERE are they located? NEO-M8N GPS + MS5837 Pressure Sensor + H3 Hex Index 3D distribution mapping (latitude, longitude, and water column depth)
3. HOW MUCH is available? Length-Weight calculation ($W = a \cdot L^b$) + Crate Volume CPUE (Catch Per Unit Effort in kg/haul-hour) and total harvest biomass
4. HOW do patterns change? TimescaleDB fusing Temp (DS18B20), Turbidity & Drift Spatiotemporal migration trends and seasonal thermocline shifts
5. WHAT does it mean? On-device voice advisor & CMFRI scientific dashboard Fuel-optimal routes for fishers; sustainable harvest limits for scientists

5. Summary Cheat-Sheet: Questions Judges Will Ask Us

  1. "Why wouldn't fishermen just use a smartphone app?"
    • Answer: Smartphones cannot measure water temperature or depth, cannot survive salt spray when dropped on deck, and fishermen with wet, salty hands cannot type while handling fishing nets. MATSYA-Node operates 100% hands-free.
  2. "Why did you choose LoRa instead of Satellite communications?"
    • Answer: Marine satellite terminals cost over ₹60,000 to buy and ₹3,000–₹8,000 per month in airtime subscriptions, making them unaffordable for small-scale motorized fishers. LoRa runs on India's license-free 865–867 MHz band with zero monthly fees.
  3. "How do you ensure fishermen don't reject the device out of privacy fears?"
    • Answer: Our Dual-Ledger protocol mathematically strips exact GPS coordinates before data leaves the boat, sharing only aggregated 8–10 km hexagonal zones with CMFRI while preserving exact spots solely on the fisherman's personal device.

Top comments (0)