Scout Drone ROS Implementation

A comprehensive ROS-based suite for drone-based perception, sensor fusion, and geolocation. This repository implements a "Scout" drone system capable of detecting targets (YOLOv8), thermal sensing, and precise ground-coordinate calculation

---

Key Modules

1. Perception & YOLO Detection

Implements a real-time object detection pipeline.

• Camera Node: Publishes raw BGR images from webcams, IP cameras, or simulated video files to /camera/image_raw.
• YOLO Detector: Subscribes to images and performs inference using YOLOv8. It publishes structured YoloDetectionArray messages containing bounding boxes and confidence scores.
• Optimization: Supports CUDA acceleration for low-latency detection on hardware like Jetson Nano.

2. Geolocation Engine

The "Mission Brain" that calculates real-world GPS coordinates of detected targets.

• Sensor Fusion: Uses ApproximateTimeSynchronizer to align YOLO detections with Drone GPS (Lat/Lon), IMU (Pitch/Roll/Yaw), and LiDAR (Altitude).
• Coordinate Projection: Projects 3D rays from the camera frame onto the ground plane using camera intrinsic parameters and trigonometric ground-intersection math.
• Global Mapping: Converts local North/East meter offsets into global Latitude/Longitude using the pyproj library.

3. Multi-Sensor Integration (Lidar/Camera/Thermal)

Core implementation for physical and simulated sensor nodes.

• High-Fidelity Simulation: Includes fallback patterns for testing without hardware:
  • Thermal: Simulated Gaussian heat spots with COLORMAP_JET visualization.
  • Camera: Animated gradient test patterns with crosshair overlays.
  • LiDAR: Hovering altitude simulation with sinusoidal noise.
• Hardware Drivers: Pre-built support for TFMini Plus (Serial), CSI cameras, and standard UVC thermal sensors.

---

🛠 Setup & Installation

Prerequisites

• ROS Noetic (Ubuntu 20.04 / WSL2)
• Python Dependencies:

  pip install ultralytics pyproj scipy opencv-python cv_bridge

Build Workspace

# Clone the repository into your catkin workspace source folder
cd ~/catkin_ws/src
# (Paste repository here)

# Build
cd ~/catkin_ws
catkin_make
source devel/setup.bash

---

How to Run

Testing with Simulation (No Hardware)

The project includes a test_perception.launch file that simulates a drone flying while running the full perception pipeline.

1. Start Master: roscore
2. Launch System: roslaunch scout_nodes test_perception.launch
3. Monitor Output:

   # View final calculated target GPS coordinates
   rostopic echo /mission/target_coordinates
   
   # View camera/thermal streams
   rosrun rqt_image_view rqt_image_view

Real Hardware Deployment

Modify the CAMERA_SOURCE or LIDAR_PORT parameters in the respective node scripts to point to your device paths (e.g., /dev/video0 or COM3).

---

Repository Structure

• camera_yolo/: YOLOv8 node and custom message definitions.
• geolocation/: The core math engine for target positioning and fake drone data publishers.
• Lidar-Camera-thermal/: Implementation of the physical/simulated sensor nodes.
• remaining.md: Design docs for future Decision and Communication nodes.

🔮 Future Roadmap

• Decision Node: Autonomous grid-search patterns.
• Comm Gateway: Secure telemetry bridge between Drone and Base Station.
• Delivery Commander: MAVLink-based payload release for the secondary drone.