Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: THEend8_
This assignment will enable you to:
• Deepen your understanding of camera calibration, keypoints detection / matching, homography, fundamental matrix, and camera pose estimation.
• Recognize software design challenges behind implementations of computer vision algorithms.
• Design and optimise software to meet specified requirements.
• Acquire a hands-on understanding of camera calibration and simultaneous localisation and mapping problems.
(These correspond to point 1, 2, 4 and 5 of the module learning outcomes. Module learning outcomes are provided in the Module Descriptor)
The assignment consists of two main tasks. The first task is to perform camera calibration using images stored in theCalibrationImages_MVO.zip file. These calibration images were captured with a checkerboard calibration pattern placed at different positions and orientations. The size of the checkerboard square is 14.44mm x 14.44mm.
The second task is to estimate three-dimensional camera poses (position & orientation) for the sequence of images from theCVML Monocular Visual Odometry dataset stored in theCVML_MVO_Loop.zip file. These images were captured with varying camera position and orientation. The images in both theCalibrationImages_MVO andCVML_MVO_Loop were taken by the same camera. You are asked to write matlab programs to estimate intrinsic camera parameters using data in theCalibrationImages_MVO.zip file and subsequently estimate the camera pose for each corresponding image in theCVML_MVO_Loop.zip sequence.
In visual odometry, an estimate of the global pose of the camera for the current frame tends to drift from the true pose due to matching errors between consecutive frames. If camera trajectory loops, shown the same part of the scene as before, this can be used to correct some of the camera pose drift errors. You to implement algorithm for such “loop closure”.
It is essential that you design your camera pose estimation algorithm, so it can be used in a sequential manner, i.e., when estimating the current camera pose only the current and preceding images can be used.
TheCalibrationImages_MVO_Loop.zip andCVML_MVO_Loop.zip files are available from Blackboard EL3105 Assignment space.
References:
Hartley, R. and Zisserman, A. (2003), Multiple View Geometry in Computer Vision, Cambridge University Press.
Szeliski, R.. (2022), Computer Vision: Algorithms and Applications”, Springer, Chapter 7 Structure from Motion (pp. 345-377).
Bay, R., Tuytelaars, T. and Gool, L.V. (2006), SURF: Speed Up Robust Features”, European Conference on Computer Vision, ECCV’2006, pp. 404-417.
Mikolajczyk, K. and Schmid, C. (2005), A performance evaluation of local descriptors, IEEE Transactions on Pattern Analysis and Machine Intelligence, Volume 27, Issue 10.
B. Triggs, et al. (2002) Bundle Adjustment – A Modern Synthesis, International Workshop on Vision Algorithms.
Matlab help:
“Monocular Visual Odometry”
“Monocular Visual Simultaneous Localization and Mapping
Late work
Work submitted electronically may be submitted after the deadline to the same Turnitin assignment slot and will be automatically flagged as late. Except where an extension of the hand-in deadline date has been approved lateness penalties will be applied in accordance with the University policy as follows:
(Working) Days |
|
Late Penalty |
1 - 5 |
|
maximum mark that can be achieved: 40% |
more than 5 |
|
0% given |