Project Outline

Accurate 3D reconstruction of crops and agricultural environments in digital agriculture brings precision, efficiency, and sustainability to farming practices. By providing detailed and accurate 3D models, it enables better monitoring, resource management, yield prediction, and overall farm management. This supports the move towards smarter, data-driven agricultural practices, ultimately enhancing productivity and sustainability in the agricultural sector.

This project aims to develop accurate, scalable, and efficient 3D reconstruction approaches using multi-view imagery for digital agriculture. Firstly, the research will explore and utilize novel 3D representations, such as neural implicit representation and 3D Gaussian Splatting, which are expected to offer significant improvements in the speed, scalability, and accuracy of 3D reconstruction. The study will evaluate their performance and applicability in large-scale agricultural settings, where fast and accurate reconstruction is crucial.  

Secondly, the study seeks to integrate expert prior knowledge of biological structure information, (e.g. functional structural plant models (FSPM)), into the reconstruction framework. By doing so, it will enforce structure-aware reconstruction, ensuring that the reconstructed 3D models accurately reflect the biological and structural characteristics of the plants. This integration will help in providing a more detailed and realistic representation of crops, enabling better monitoring and management in digital agriculture.  

Finally, this study will develop methods to account for and mitigate the effects of noisy camera poses, thereby enhancing the robustness and reliability of the 3D reconstruction process. This will ensure that the models are accurate and useful even in less-than-ideal data collection conditions such as using drone to capture image in large scale.

The outcomes of this research are expected to provide promising economic and environmental benefits and lay the groundwork for future innovations in digital agriculture.

You will:

• Develop advanced understanding of 3D Reconstruction using multi-view images; in-depth knowledge of state-of-the-art 3D reconstruction methods; and learn how to implement and evaluate the performance of these methods in real-world agricultural settings;
• Learn to incorporate functional structural plant models (FSPM) into technological frameworks, ensuring that biological and structural characteristics are accurately represented in 3D reconstructions, which can help us understand the significance of biological constraints in enhancing the accuracy and utility of digital models in agriculture.
• Gain practical experience in digital agriculture through hands-on experience in collecting and processing agricultural data using drones and other sensors. Further, the student will understand the practical challenges and solutions in implementing advanced digital technologies in agricultural environments.
• Be engaged in collaborative work with experts at ANU and CSIRO, including crop breeders, biologists, and computer vision experts. You will also have opportunities to directly engage with some of CSIRO’s existing industry partners. These interactions will provide you with valuable interdisciplinary research experience, and a better understanding of industry to prepare you to tackle real-world problems effectively in your future career.
• Develop research and communication skills through comprehensive literature review, experimental design, data analysis, interpretation of results, academic paper writing and participation in conferences and other relevant events and trainings.
• Emerge with a well-rounded skill set, combining technical expertise with practical experience and research capabilities, positioning you for a successful career in digital agriculture and related fields.

Supervisors:

• Xuesong Li, Research Scientist in 3D computer vision, vision–language models (VLMs), and robotics, CSIRO
• Jing Zhang, Senior Lecturer in computer vision, machine learning and generative AI, School of Computing, ANU

Questions about this project can be directed to:  Xuesong.li@csiro.au and  jing.zhang@anu.edu.au

To register an expression of interest: click here. You will need to outline why you have selected the research project and how your skills, experience and/or knowledge meet the project requirements.