Random Sample Consensus
(Added more text about RANSAC) |
(Added RANSAC video) |
||
| Line 1: | Line 1: | ||
{|align="right" | {|align="right" | ||
|- | |- | ||
| − | |[[Image:Rodriguez.jpg|thumb| | + | |<html> |
| + | <div class="thumb tright"> | ||
| + | <div style="width:320px;"> | ||
| + | <embed style="width:320px; height:240px;" id="VideoPlayback" type="application/x-shockwave-flash" | ||
| + | src="http://vision.eng.shu.ac.uk/jan/flv/flvplayer.swf" width="320" height="240" flashvars="file=http://vision.eng.shu.ac.uk/jan/flv/ransac.flv&image=http://vision.eng.shu.ac.uk/jan/flv/ransac.jpg&displayheight=240" | ||
| + | pluginspage="http://www.macromedia.com/go/getflashplayer"/> | ||
| + | <div class="thumbcaption" > | ||
| + | 2D rigid object recognition with RANSAC. In this case appearance templates where used as feature descriptors | ||
| + | </div> | ||
| + | </div> | ||
| + | </div> | ||
| + | </html> | ||
| + | |- | ||
| + | |[[Image:Rodriguez.jpg|thumb|240px|The RANSAC implementation was done by ''Sonia Fernández RodrÃguez'', ERASMUS exchange student from Spain]] | ||
|- | |- | ||
|} | |} | ||
=RANSAC= | =RANSAC= | ||
| − | '''RANSAC''' is a well-known algorithm for dealing with outliers while fitting a parametric model to some input data. Applied to object recognition RANSAC requires scene-features, model-features, and a similarity measure as input. A randomly selected set of features (one feature for translational model, two features for estimating translation and rotation) in the scene is picked and a fit for the pose of the model is generated which maps the best correspondences of the model into the scene. If the geometric constraints are fulfilled, the number of inliers is computed (i.e. the number of model features having a good correspondence at the appropriate location in the scene). This final step is called the consensus step. If the number of correspondences is considered to be sufficiently high, this larger set of correspondences is used to improve the parameters of the final estimate. See [http://vision.eng.shu.ac.uk/jan/report_rodriguez.pdf project report] for more details. | + | '''RANSAC''' is a well-known algorithm for dealing with outliers while fitting a parametric model to some input data. Applied to object recognition RANSAC requires scene-features, model-features, and feature descriptors with a similarity measure as input. A randomly selected set of features (one feature for translational model, two features for estimating translation and rotation) in the scene is picked and a fit for the pose of the model is generated which maps the best correspondences of the model into the scene. If the geometric constraints are fulfilled, the number of inliers is computed (i.e. the number of model features having a good correspondence at the appropriate location in the scene). This final step is called the consensus step. If the number of correspondences is considered to be sufficiently high, this larger set of correspondences is used to improve the parameters of the final estimate. See [http://vision.eng.shu.ac.uk/jan/report_rodriguez.pdf project report] for more details. The best results where achieved by using appearance templates as feature descriptors [http://www.cs.bris.ac.uk/home/pupilli/publications/thesis.pdf]. |
=See Also= | =See Also= | ||
| Line 13: | Line 26: | ||
* [http://vision.eng.shu.ac.uk/jan/report_rodriguez.pdf Recognition of 2D-objects using RANSAC] | * [http://vision.eng.shu.ac.uk/jan/report_rodriguez.pdf Recognition of 2D-objects using RANSAC] | ||
* [http://en.wikipedia.org/wiki/RANSAC Wikipedia article on RANSAC] | * [http://en.wikipedia.org/wiki/RANSAC Wikipedia article on RANSAC] | ||
| + | * [http://www.cs.bris.ac.uk/home/pupilli/publications/thesis.pdf Particle Filtering for Real-time Camera Localisation] | ||
Revision as of 23:49, 28 May 2008
|
2D rigid object recognition with RANSAC. In this case appearance templates where used as feature descriptors
|
 |
RANSAC
RANSAC is a well-known algorithm for dealing with outliers while fitting a parametric model to some input data. Applied to object recognition RANSAC requires scene-features, model-features, and feature descriptors with a similarity measure as input. A randomly selected set of features (one feature for translational model, two features for estimating translation and rotation) in the scene is picked and a fit for the pose of the model is generated which maps the best correspondences of the model into the scene. If the geometric constraints are fulfilled, the number of inliers is computed (i.e. the number of model features having a good correspondence at the appropriate location in the scene). This final step is called the consensus step. If the number of correspondences is considered to be sufficiently high, this larger set of correspondences is used to improve the parameters of the final estimate. See project report for more details. The best results where achieved by using appearance templates as feature descriptors [1].
