1 / 7

REU Project

REU Project. 4D Efficient Real-Time Graph Based Temporal Point Cloud Segmentation. Steven Hickson. Current Method. Pros: Fast, currently at 400 ms / frame Can be run in real time (especially if we convert some code to CUDA) Has no memory or video length limit, ie can run infinitely.

jalena
Télécharger la présentation

REU Project

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. REU Project 4D Efficient Real-Time Graph Based Temporal Point Cloud Segmentation Steven Hickson

  2. Current Method Pros: • Fast, currently at 400 ms / frame • Can be run in real time (especially if we convert some code to CUDA) • Has no memory or video length limit, ie can run infinitely. Cons: • Still not as accurate. • Re-labels segments if there is too much occlusion. • Still has 5 arbitrary values that determine the segmentation. • Over-reliant on the centroid, (this can be easily changed but needs to be experimented on.

  3. Hierarchal Region Trees • Whereas Georgia tech used Region Graphs, we use Region Trees, which are constructed using the labeled graph combined with the original point cloud data. • These are constructed with only one level, however, the tree can be made hierarchal by merging the tree upwards based off the LABD histogram difference between each region and its pre-computed neighbors.

  4. Basic Region Tree

  5. Propagated Region Tree

  6. Benefits • No more arbitrary values. Only one user input, which determines which tree level is selected. • Better segmentation since the histogram of the region leads to a better estimation. • Robust and novel approach

  7. Results Original 85% Level 65% Level 45% Level

More Related