Presentations

Three different kinds assumptions are often used for representing the shape of a paper:rigid, foldable and nonrigid. Nonrigid surface detection is intensively explored as challenging topics which addresses two problems: recovering the paper shape and estimating the camera pose. The state-of-the-art researches try to solve both problems robustly for real time purpose. We propose an augmented reality application that use a nonrigid detection method to recover the shape of the bendable paper using dots as keypoints and estimate the camera pose simultaneously. Our approach recovers the multi-planarity of the paper as the initial shape and iteratively approximates the surface shape. The multi-planarity is estimated by using the tracking by descriptor update method that uses the correspondence between captured and reference keypoints. We then optimize the shape using the progressive finite newton optimization method.

Raster maps are widely available in the everyday life, and can contain a huge amount of information of any kind using labels, pictograms, or color code e.g. But due to those overlapping features, it’s not an easy task to extract roads from those maps. Many methods use user input to achieve this goal. In this paper we focus on an automated method to extract roads by using color segmentation and linear features detection to search for seed points having a high probability to belong to roads. Those seeds are then expanded before choosing to keep or to discard the extracted element.