Unclassified // Public Domain // Academic Research Project
Defense Technology
Project Designation

S.A.R.D

Surveillance And Reconnaissance Drone

Max Thrust
31.4
Newtons
Range
1.9
Kilometers
Detection
~100
Faces/Frame
Thrust/Weight
2.4:1
Ratio
01 // MISSION BRIEF

Project Overview

Tactical quadcopter for surveillance and reconnaissance operations

S.A.R.D is a tactical quadcopter designed for surveillance and reconnaissance missions, engineered to meet operational requirements of military and security forces. The platform provides real-time target tracking, high-resolution optics, and serves as a cost-effective alternative to military-grade UAV systems.

The system integrates GPS navigation, a 360-degree dual camera module, and facial recognition algorithms powered by YuNet Deep Neural Network. This configuration enables real-time target identification with detection of approximately 100 faces per frame and high-precision recognition at 2-meter range, with wide-area aerial reconnaissance capabilities extending to 50 meters.

Architected with open system design principles, S.A.R.D features rapid battery swap capability, dual-camera surveillance coverage, on-edge AI/ML processing, vibration-dampening hardware, and 8GB storage for mission data recording. Total development cost maintained at $476, demonstrating force multiplication through accessible technology.

02 // DEVELOPMENT TIMELINE

Build Process

Complete development cycle from concept to deployment

Phase 01

Design & Planning

Initial system architecture design, component selection, and mission requirements definition. Conducted feasibility analysis for integrating AI-powered facial recognition with quadcopter platform.

CAD modeling of pan-tilt camera assembly in OnShape
Component selection: KK2.1.5 flight controller, Raspberry Pi 4B, Arduino Uno
System architecture design for dual-controller avionics suite
Power budget calculations and battery selection
Phase 02

Parts Procurement

Ordered all hardware components totaling $476, including airframe, motors, flight controller, computing modules, sensors, and camera systems. Prioritized cost-effective solutions while maintaining performance requirements.

Sourced DXW multi-rotor airframe with 500mm wheelbase
Acquired Arducam 64MP camera module and OV2659 2MP USB camera
Purchased LSM9DS1 9-axis IMU and NEO-6M GPS module
Selected RadioLink T8FB 8-channel transmitter for RC control
Phase 03

Hardware Integration

Physical assembly of airframe, motor mounts, flight controller installation, and initial power distribution. 3D-printed custom pan-tilt mount for camera gimbal system.

Assembled quadcopter frame with motor mounts and propellers
Integrated KK2.1.5 flight controller with ESCs and motors
3D-printed pan-tilt camera assembly and mounted servos
Installed vibration dampers for camera stabilization
Phase 04

PID Controller Tuning

Configured PID gains for flight stabilization through KK2.1.5 interface. Tuned proportional, integral, and derivative parameters for stable hover and responsive control.

Initial P-gain tuning for pitch, roll, and yaw axes
I-gain adjustment to eliminate steady-state error
D-gain optimization to reduce oscillations
Connected Raspberry Pi GPIO pins to servo motors for pan-tilt control
Phase 05

Flight Controller Testing

Conducted ground tests and initial flight tests to validate flight controller performance, motor response, and overall system stability before adding payload complexity.

Ground testing: motor spin tests, RC signal validation
Hover tests with incremental altitude increases
Stability assessment in various flight modes
Measured thrust-to-weight ratio: 2.4:1
Phase 06

Arduino IMU & GPS Programming

Developed C++ firmware for Arduino Uno to interface with LSM9DS1 9-axis IMU and NEO-6M GPS module. Implemented I2C communication protocols and data parsing algorithms.

I2C library implementation for LSM9DS1 accelerometer/gyroscope
Serial communication setup for NEO-6M GPS (NMEA parsing)
Data fusion algorithm combining IMU and GPS readings
Achieved GPS position accuracy of 2.5m CEP
Phase 07

AI Camera Programming

Implemented YuNet Deep Neural Network on Raspberry Pi 4B for real-time facial detection and tracking. Developed Python interface using OpenCV for camera input and computer vision processing.

YuNet model deployment on Raspberry Pi with OpenCV
Real-time video stream processing from 64MP Arducam
Facial detection achieving ~100 faces per frame
Pan-tilt servo control for automatic target tracking
Phase 08

Sensor Data Collection

Collected flight data from IMU, GPS, and camera systems during test flights. Logged telemetry data for post-flight analysis and algorithm refinement.

Recorded 9-axis IMU data at 100Hz sampling rate
GPS position logging with timestamp synchronization
Video recording with facial detection overlays
Compiled 8GB of flight data for analysis
Phase 09

Kalman Filter Implementation

Implemented Kalman filter algorithm in Python to process noisy IMU sensor data. Fused accelerometer, gyroscope, and magnetometer readings for improved attitude estimation.

Extended Kalman Filter (EKF) for sensor fusion
State prediction using IMU acceleration/gyro data
Measurement update using GPS position
Reduced position estimation error by 40%
Phase 10

Integration & Testing

Final system integration, comprehensive testing, and demonstration at Florida Tech Engineering Showcase. Validated all subsystems working cohesively.

Full system integration test with all subsystems active
Live demonstration of facial tracking capabilities
Achieved 7-10 minute flight endurance with full payload
Successful public demonstration at FIT Engineering Expo
03 // CAPABILITIES

System Capabilities

Operational features and technical achievements

AI

Neural Target Tracking

Real-time facial recognition using YuNet Deep Neural Network on 64MP Arducam sensor. Detection of ~100 targets per frame with high-precision identification at 2m and wide-area reconnaissance to 50m range.

SYS

Dual-Controller Architecture

Integrated avionics suite: KK2.1.5 flight controller for stabilization, Arduino Uno for sensor data acquisition (9-Axis IMU, GPS), Raspberry Pi 4B for image processing and servo control of pan-tilt mechanism.

OPT

360° Surveillance System

Custom-designed pan-tilt camera assembly with 2-axis servo control for continuous target tracking. CAD-modeled and 3D-printed OnShape assembly with integrated micro-servo mounts and Arducam housing.

FLT

Optimized Flight Performance

PID controller tuning via GPIO interface from Raspberry Pi to servo actuators. IMU calibration with Kalman filter achieving 7-10 minute flight endurance, 2.4:1 thrust-to-weight ratio, GPS accuracy of 2.5m.

04 // SPECIFICATIONS

Technical Parameters

Performance metrics and operational limits

System Weight
1.98
kg
Endurance
7-10
minutes
Max Thrust
31.4
N
Range
1.9
km
T/W Ratio
2.4:1
ratio
GPS Accuracy
2.5
m CEP
05 // VISUAL DOCUMENTATION

Project Gallery

Development, testing, and operational demonstration

06 // TECHNICAL COMPETENCIES

Skills Demonstrated

Engineering disciplines and technologies applied

Systems Integration
CAD & Additive Manufacturing
Python Development
C++ Embedded Programming
Sensor Fusion
PID Control Systems
Computer Vision (OpenCV)
Embedded Linux (Raspberry Pi)
Hardware Prototyping
Flight Dynamics
Kalman Filtering
Real-Time Systems
I2C/Serial Communication
Deep Learning (YuNet)
Data Acquisition Systems
Avionics Integration
07 // BILL OF MATERIALS

System Components

Complete hardware breakdown and cost analysis

Power & Airframe
Zeee 3S LiPo Battery 2200mAh QTY: 2 $34.99
LiPo Battery Charger QTY: 1 $14.94
DXW Multi-Rotor Airframe QTY: 1 $155.99
Propeller Set (12-Pack) QTY: 1 $8.29
Control Systems
Raspberry Pi 4B 4GB RAM $64.45
KK2.1.5 Flight Controller QTY: 1 $15.69
Arduino Nano Board QTY: 1 $11.99
MG90S Micro Servo Motor 9G $13.99
Sensors & Telemetry
LSM9DS1 9-Axis IMU QTY: 1 $16.39
NEO-6M GPS Module QTY: 1 $12.59
RadioLink T8FB 8CH Transmitter $59.99
Optical Systems
Pilipane 2MP Camera (OV2659) USB $10.70
Arducam 64MP Camera QTY: 1 $55.99
Total System Cost $476.00
Next
Next

CFD analysis of my car