# Matlab代写 | Computer Vision and Imaging [06-30213] Summative Assignment 2

### Matlab代写 | Computer Vision and Imaging [06-30213] Summative Assignment 2

Instructions
This assessment is summative and contains two parts. In Part 1, you will carry out object/scene
reconstruction with Structure from Motion. In Part 2, you will implement principal component
analysis in MATLAB for two di erent data sets, along with designing your own facial recognition
tool.
Your answer must be submitted to Canvas before the deadline in the form of a single zip archive
le containing:
1. Your answers to the questions in prose and diagrams. This should take the form of a single
PDF document with the answers for each question using the provided LaTeX template.
2. Your code and any accompanying les necessary to execute the code for any programming
questions as speci ed in the LaTeX template.
and a separate PDF document with the answers for Turnitin checking (two les in total; one zip
le and one PDF le). Some or all of the text of each question is emphasised using italics. This
emphasis indicates a question that must be explicitly answered or a task that must be completed.

Part 1
Structure from Motion involves reconstructing the 3D structure of a scene from multiple images.
Given the size of an object in the scene, it is possible to recover the actual sizes of everything in
the scene given sucient image quality.
Question 1.1 There are three parts to Question 1.1. The aim of this question is to recreate dense
reconstructions of objects from two images.
Question 1.1.1 [4 marks] However, you might get the situation where the reconstruction is
more two dimensional than three dimensional. Can you state a reason why that is and
two ways to rectify it? You might need to run your solutions to the following sub-parts
of Part 1 multiple times to achieve a result you are satis ed with.
Summary of what needs to be answered:
1. One reason why a reconstruction is more two-dimensional (report in PDF)
2. Two ways to rectify the above (report in PDF)
Your solution to the following questions for this Part 1 (1.1, 1.2 and 1.3) should:
1. Use any pre-processing you like to manipulate the given images
2. A series of chessboard images are provided. Use them to generate a calibration le for
the following questions using the camera calibrator app in MATLAB. The chessboard
squares are 20mm a side. Use this le to remove any distortion from the given images.
3. Detect features using the minimum eigenvalue algorithm
4. Use a point tracker to track the matched features
5. Estimate the fundamental matrix from the matched points with any method, such as
MSAC or RANSAC etc., and check its validity
6. Triangulate over the matched points
7. Construct and display a point cloud
8. Obtain the colour values for the point cloud
9. The reconstruction has to be a dense reconstruction
Question 1.1.2 [8 marks] Two images of the University Library are provided. Reconstruct
the library. Include an image of the tracked features and two images of your coloured
reconstruction, showing it at di erent angles. Include an image of the relative camera
positions after calibration as well as the radial distortion and mean re-projection error
in your report. When running the MATLAB le, these images (not including the cal-
ibration) should all appear as gures and the reconstruction as a movable graph. Save
and include the MATLAB reconstruction as a . g le and the calibration .mat le in
Summary of what needs to be answered:
1. Image of relative camera positions (report in PDF)