Part 1 – Tracking Texture

Our project has decided to explore four different approaches as described in an earlier post in this blog. My project partner is focusing on Shape Based & Colour Based tracking and I am going to focus my efforts on tracking which focuses on motion and texture.

Kalman Filtering

The first thing I notice while researching this topic is that Kalman filtering seems to be a popular algorithm when it comes to this approach.

Kalman filtering has a variety of applications and like all filters it allows certain things to pass through and other things not. The Kalman filter’s aim is to filter imperfect information, sort out the useful parts of interest and reduce the uncertainty and noise.

Apparently an early application of Kalman filtering was used for guided missiles and was also used as part of the onboard navigation system aboard the Apollo 11 Lunar module.

Particle Filtering

Unlike the Kalman Filter, particle filtering takes a non-linear approach. Particle filtering in a nutshell involves representing a posterior function by a set of random samples (particles).

Initial research into the area of texture tracking seems to suggest that it’s effectiveness is dependent on the image in question possessing strong textures (unsurprisingly) and texture tracking techniques tend to get used in conjunction with other means of detection and tracking.

The series of images we are working with in this project is neither rich in texture or indeed detail as the image is quite limited in terms of resolution.

So with that in mind, I am going to park this avenue of research here for now and may revisit it later should I decide to use this technique in combination with another approach

OpenCV offers its own Kalman filter function, the constructor of which I have included below.

REFERENCES:

Pressigout M, Marchand E. Real-Time Hybrid Tracking using Edge and Texture Information. The International Journal of Robotics Research Vol 26, No 7, July 2007
http://www.irisa.fr/lagadic/pdf/2007_ijrr_pressigout.pdf

Particle filter (accessed 25.10 2018)
https://en.wikipedia.org/wiki/Particle_filter

Motion Analysis and Object Tracking (accessed 25.10 2018)
https://docs.opencv.org/3.0-beta/modules/video/doc/motion_analysis_and_object_tracking.html

Tracking – Tracking by Background Subtraction (accessed 25.10 2018)
https://www.hdm-stuttgart.de/~maucher/Python/ComputerVision/html/Tracking.html

Vacchetti L. Lepetit V. Fua P. Combining Edge and Texture Information
for Real-Time Accurate 3D Camera Tracking. CVlab. Swiss Federal Institute of Technology, 1015 Lausanne, Switzerland
https://www.labri.fr/perso/vlepetit/pubs/vacchetti_ismar04.pdf