Dual-Tech UGV Drony Na Odlew

Overview

This project implements a reconnaissance system for an unmanned ground vehicle (UGV) running on Raspberry Pi. It supports both manual and autonomous driving modes, using a camera, GPS, QR code recognition, and an AI model (YOLOv8) to detect and classify military and civilian objects.

Main features:

• Real-time keyboard control
• Movement tracking and route replay
• QR code and associated image recognition
• YOLO-based object detection and classification
• Automated mission report generation

Directory Structure

umarlypoeta-dual_tech_konkurs_kod/
├── main.py / main2.py               # Basic control scripts
├── vehicle_control_server.py        # Flask server for remote control
├── start_server.sh                  # Script to launch the server
├── raport.txt / raport_misji.txt    # Example reports
├── architektura/                    # Legacy modules
│   ├── camera.py, engine.py, gps.py, start.py
└── wersja_nowa/                     # Final implementation
    ├── main_ost.py                  # Final version with threading & AI
    ├── main2_auto*.py               # Alternative auto-driving scripts
    ├── create_file.py               # Mission report generator
    ├── robienie_zdjec*.py           # Image dataset creation scripts
    └── ai/
        └── categorize.py            # ONNX model image categorizer

Requirements

• Raspberry Pi (tested on Pi 5)
• Python 3.7+
• Dependencies:

pip3 install flask RPi.GPIO gpiozero opencv-python pyzbar picamera2 onnxruntime numpy

• Compatible camera (Picamera2)
• GPS module (connected via UART)
• Trained YOLOv8 model (best.onnx)

Usage

Manual & Auto Modes

Run the main script:

python3 wersja_nowa/main_ost.py

Manual Mode Controls:

In Auto Mode, the UGV replays the saved movement path from track_log.json.

Features

✅ QR Code Recognition

• Scans and logs QR codes using Pyzbar
• Extracts and compares images located beneath QR codes via OpenCV

✅ GPS Integration

• Retrieves and logs real-time coordinates
• GPS logs are included in all object and QR detections

✅ Object Detection (YOLOv8)

• Detects predefined object types using best.onnx
• Detected objects are classified and saved in rozpoznane_obiekty.txt

✅ Mission Report Generation

Run:

python3 wersja_nowa/create_file.py

It generates raport_misji.txt including:

• Team name
• Recognized objects
• Traversed and optimal paths
• Length of optimal route

Dataset Creation for Training

Run:

python3 wersja_nowa/robienie_zdjec3.py

• Captures images from the camera
• Saves them into YOLO-compatible folder structure
• Generates empty label files for annotation

Offline Image Categorization

Run:

python3 wersja_nowa/ai/categorize.py

• Processes all images in images/
• Classifies each using the ONNX model
• Saves results to results.txt

Author

umarlypoeta
Entry for Dual-Tech Challenge
📅 April 2025