7CCEMSAP Coursework 1
Coursework 1: 3D Reconstruction for Robotic Manipulation
7CCEMSAP Sensing & Perception
• This coursework will count for 30% of your total grade in the course.
• Remember that all work you submit should be your own work. The College treats plagiarism very seriously.
• You should submit your report by 20th November 2025 using the submission link on the KEATS site for 7CCEMSAP.
• If you have more general or administrative problems please use the online forum or consult during drop-in-sessions or support sessions. If an email is required, please include the course number (7CCEMSAP) in the subject line.
1 Objective
Develop an understanding of 3D reconstruction techniques in computer vision and apply these tech- niques to a practical robotics context, specifically focusing on manipulation. Students will implement a system that constructs a 3D model from a set of 2D images and then utilise this model to guide a robotic arm for manipulation. This project will provide hands-on experience in both 3D computer vision and robotics.
2 Description
3D Reconstruction:
1. Image Capture: Capture a series of 2D images of an object from multiple angles using a standard camera.
2. Feature Detection and Matching: Implement a feature detection algorithm to identify key points in each image and match these features across the image set. You can use computer vision libraries such as OpenCV.
3. Camera Calibration and Pose Estimation: Determine the camera parameters and estimate its pose relative to the object in each image.
4. Triangulation: Use matched feature points and camera parameters to triangulate 3D points and create a sparse 3D reconstruction of the object.
5. Dense Reconstruction (Optional): Enhance the sparse model to a dense 3D reconstruction using techniques like Multi-View Stereo. You can use computer vision libraries such as PCL.
Robotic Application:
1. Integration with Robotic Simulator: Import the 3D model into a robotic simulation environment (such as Gazebo or ROS).
2. Grasping Algorithm: Implement an algorithm that enables a robotic arm in the simulator to identify potential grasping points on the 3D model.
3. Manipulation: Program the robotic arm to perform simple manipulation tasks using the identified grasping points, such as picking up and moving the object.
3 Deliverables
1. A technical report summarising the methodology, challenges faced, and main findings (maximum 4 pages, including the references).
2. The source code for the 3D reconstruction and robotic manipulation tasks. Please include a READ ME file, and make sure your code can be run easily, so that we can reproduce your results.
3. A short video (maximum 2 minutes) demonstrating the working system, and highlighting the key aspect of the implementation.