• gaussian
• filter
• mixture
• camera
• stitching
• kalman

Algorithms for aligning images and stitching them into seamless photo- mosaics are among the oldest and most widely used in computer vision. Image stitching algorithms create the high-resolution photo-mosaics used to produce today’s digital maps and satellite photos. They also come bundled with most digital cameras and can be used to create beautiful ultra wide-angle panoramas.
The scope of this thesis is to write a robust algorithm that reconstruct the geometry of a large moving vehicle that passes in front of a fixed camera capturing frames at fixed frequency. The camera is arranged in a portal structure embedding among the camera, a laser and fixed illumination source.
The target object whose geometry has to be estimated is a train traveling through the portal at a variable velocity.
State of the Art stitching algorithms are not suitable for the case study due to the portal structure, i.e. the fixed illumination combined with the motion of the train makes the luminance of the pixels variable over the time and then makes that algorithm frail.
Therefore, the purpose of this thesis is to combine some methods presents on the State of the art with other algorithm like the Extended Kalman Filter and the Expectation Maximization for Gaussian Mixture Model in order to make the overall system robust to this type of problems.