Crop Row Detection and Waypoint Definition Using Aerial Images of Fields

Authors

Alireza Amiri
Department of Mechatronics Engineering
Khaje Nasir Toosi University of Technology (KNTU), Tehran, Iran
ali.amiri@email.kntu.ac.ir

Abstract

As autonomous agriculture evolves, precise waypoint generation is crucial for defining the paths of autonomous robots. This project presents a method leveraging high-resolution aerial imagery to detect crop positions and determine waypoints. The approach includes:

• Hardware Setup: A camera, wireless transmitter, and receiver for capturing live images.
• Image Processing: Detecting crops using Unet and K-means clustering; applying Hough Transform for detecting crop row lines.
• Waypoint Generation: Identifying points along crop rows and converting these into global coordinates for real-time navigation.

Introduction

In the quest to enhance agricultural productivity and sustainability, autonomous robots play a pivotal role. This project addresses the challenge of accurately defining paths for these robots by integrating advanced image processing techniques and machine learning models. Traditional GPS-based navigation systems face challenges in precise row alignment, which this project aims to overcome using UAV-based aerial imagery.

Methodology

1. Data Acquisition

• Camera Selection: The GoPro Hero 4 was chosen for its high-resolution RGB imaging capabilities.
• GPS Module: The Ublox-Neo-6m provides global coordinates, essential for converting local pixel positions into global coordinates.

2. Image Processing

• Preprocessing: Aerial images are divided into smaller, manageable sections.
• Crop Detection:
  • Color Filtering: Initially used but refined through HSV conversion.
  • Unet-Based Segmentation: Achieved 96% accuracy in detecting crop areas.
• Crop Row Detection:
  • Hough Transform: Applied to detect and refine lines representing crop rows.

3. Path and Waypoint Generation

• Waypoint Definition: Paths are defined as lines parallel to crop rows, with points recorded for navigation.
• Coordination Conversion:
  • Pixel to Meter: Conversion based on camera height and image length.
  • Meter to Global: Translation into global coordinates.

Results

The system successfully detects crop rows and generates waypoints with high accuracy, demonstrating its effectiveness in real-world agricultural scenarios.

Conclusion

This approach provides a robust solution for autonomous agricultural systems, combining aerial imagery with advanced image processing to ensure precise navigation. It supports autonomous robots in navigating fields efficiently and effectively.

License

This project is licensed under the MIT License - see the LICENSE file for details.