Research & development - Wageningen | More than two weeks ago
Digital Orchard Program: Implementing a Digital Twin of an Orchard
In our Digital Orchard program, we aim to create a digital twin of an orchard by combining LiDAR and common (RGB) cameras data into a “colored point cloud”. The goal is to use existing data and run new measurement campaigns allowing us to follow fruit tree development in the orchard over time. The ultimate objective is to perform instance segmentation on the tree point cloud for recognition of age of branches on apple trees in the orchard, which can be a vital step in Orchard Management around tree monitoring for the future. However, depending on the complexity instance segmentation of the trees can be a first step.
Literature Review of Related Solution and Technology
The first task is to conduct a literature review of related solutions and technologies. This will involve researching existing digital twin implementations, point cloud processing techniques, and deep learning algorithms for instance segmentation. The literature review will help us to identify the best practices and state-of-the-art techniques for implementing our digital twin.
Plan Containing a Timeline with Deliverables
The second task is to prepare a plan containing a timeline with deliverables. This plan will outline the milestones and deadlines for the project, including the literature review, measurement campaigns, point cloud processing, deep learning algorithm development, and evaluation. The plan will help us stay on track and ensure that we meet our objectives on time.
Actively Participate in and Plan Measurement Campaigns
The third task is to actively participate and plan measurement campaigns. This will involve coordinating with the team to schedule measurement campaigns, setting up recording devices, and ensuring that the data is collected accurately. The measurement campaigns will provide us with the data we need to create the digital twin and train the deep learning algorithm.
Process and Annotate Point Clouds (Using Change Detection) Making Them Suitable for Deep Learning
The fourth task is to process and annotate (colored) point clouds using change detection, making them suitable for deep learning. This will involve developing algorithms to process the point cloud data, segment the trees, and annotate the data for training the deep learning algorithm. Change detection between multiple years of data will allow us to retrieve annotation without performing much manual annotation.
Create a Deep Learning Point Cloud Algorithm and Evaluate the Deep Learning Solution
The fifth and final task is to create a deep learning point cloud algorithm and evaluate the deep learning solution. This will involve developing a deep learning algorithm for instance segmentation, training the algorithm on the annotated data, and evaluating the performance of the algorithm. The evaluation will help us determine the accuracy and effectiveness of the algorithm and identify areas for improvement.
Does this project sound like an interesting next step in your career at imec? Don’t hesitate to submit your application by clicking on ‘APPLY NOW’.
Should you have more questions about the project and the recruitment process, you can contact Martijn Kohl. He will be happy to assist you.