Feature-based object recognition and localization in 3D-space, using a single video image
Computer Vision and Image Understanding
Evaluation of Interest Point Detectors
International Journal of Computer Vision - Special issue on a special section on visual surveillance
Scale-invariant shape features for recognition of object categories
CVPR'04 Proceedings of the 2004 IEEE computer society conference on Computer vision and pattern recognition
| Hi-index | 0.00 |
We discuss a novel method of visual detection of rigid objects (that are known to the system) randomly located in complex cluttered scenes (possibly in a turbid medium that additionally deteriorates visibility). The images are captured using a gated imaging system so that (1) the medium backscattering noise is minimized and (2) only the scenes components within a predefined distance from the camera are captured in the images while the rest of the scene (containing presumably objects that are not of immediate interest) remains invisible. The database of known objects is built (in reference scale) using local shape features (interest points) extracted from template images showing the objects of interest from various viewpoints. By matching interest point detected in gates images (relative scale is used there) to the local features from the database, known objects can be identified (even if only partially visible). The paper briefly discusses the proposed methodology and explains why complexity of vision-based navigation algorithms could be dramatically reduced (while the robustness is equally dramatically improved) by adopting the proposed approach.