Colloquium: Frerik Andriessen (SSE)
24 januari 2018 13:00 - Locatie: Lecture room A, Faculty of Aerospace Engineering, Kluyverweg 1, Delft.
Ready for Detection; Stair-detection in depth images using spatial features and Adaboosting
The stairdetection algorithm is based upon recurring physical properties of staircases, and their spatial consistency.
Looking at a staircase from the front, these distances between each step become warped, as they are further away from the observer. This exponential spatial distortion is known as a 'chirp'. It is hypothesized that if an image contains a structure that matches a chirp it can be concluded to be a staircase. 20 random lines are drawn through each image, looking for a chirp. The classifications for each line are then boosted using adaboosting to come to a final classification per image.