Hand detection, segmentation, and pose recognition are challenging problems in Computer Vision with a wide variety of potential applications like alternative input devices, surveillance, motion capture, and augmented reality. This work proposes methods to solve each of these problems in high-resolution, monochromatic images via shape and texture-based methods. Hand Detection and Segmentation: This method of hand detection is based upon the outputs from both line-finding and curve-finding algorithms to find shapes that appear to be finger-like. A series of tests is performed on each finger candidate to further remove false positives and determine which sets of them could possibly form a human hand. Pose Recognition: Pose recognition works on database model, taking as input both a test image and a database of all possible hand poses. By using a scoring system comprised of votes between two different distance measures, the algorithm returns a list of database images in order of similarity to the test image. Pose Detection in a Video: To determine if a given hand pose occurs in the frames of a video sequence, the algorithm performs the pose recognition method described above, but inputs the pose to look for as the test image and the video sequence as the "database." It then returns a list of frames in the sequence in order of similarity to the given pose.
| Identifer | oai:union.ndltd.org:ucf.edu/oai:stars.library.ucf.edu:honorstheses1990-2015-1555 |
| Date | 01 January 2006 |
| Creators | Schwarz, Christopher |
| Publisher | STARS |
| Source Sets | University of Central Florida |
| Language | English |
| Detected Language | English |
| Type | text |
| Source | HIM 1990-2015 |
Page generated in 0.0086 seconds