Saliency, Scale and Image Description
International Journal of Computer Vision
Unseeded region growing for 3D image segmentation
VIP '00 Selected papers from the Pan-Sydney workshop on Visualisation - Volume 2
Robust analysis of feature spaces: color image segmentation
CVPR '97 Proceedings of the 1997 Conference on Computer Vision and Pattern Recognition (CVPR '97)
A Time-Of-Flight Depth Sensor - System Description, Issues and Solutions
CVPRW '04 Proceedings of the 2004 Conference on Computer Vision and Pattern Recognition Workshop (CVPRW'04) Volume 3 - Volume 03
Bottom-up/Top-Down Image Parsing by Attribute Graph Grammar
ICCV '05 Proceedings of the Tenth IEEE International Conference on Computer Vision - Volume 2
Recognition of Composite Human Activities through Context-Free Grammar Based Representation
CVPR '06 Proceedings of the 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition - Volume 2
Learning context-free grammars using tabular representations
Pattern Recognition
Extraction and tracking of surfaces in range image sequences
3DIM'99 Proceedings of the 2nd international conference on 3-D digital imaging and modeling
Multiple 3D sensor views object models correspondence
ICS'09 Proceedings of the 13th WSEAS international conference on Systems
| Hi-index | 0.00 |
The paper presents a perception oriented linguistic formal model for 3D sensors depth images. Both the static object extraction and short term dynamic evolution in the scene are analyzed. The target applications are subsystems in action involved in independent environment exploration and learning be it human or machine. Field of view depth images obtained with recently developed CMOS 3D sensors are analyzed in their capacity to provide immediate action oriented data. For the 3D scene images a selective segmented method is proposed in terms of salient objects in the depth image. The model as proposed uses a representation of the scene depth image in terms of object area and mean center location. An original abstract formal language representation is proposed. The extension of the context free grammar with attributes ads structure to the model. It is also shown that the generated language translates directly depth labeling into action planning on the environment. The performance of the proposed abstract representation method is analyzed in terms of estimated computation time and direct semantic relevance for a sample application. For applications of object motion detection and tracking the formal model was extended with attributes for direction and speed. The object position drift based on segment correspondence for speed determination is shown to be compatible to the formal model as proposed. Further development of the model for multi layered representations for more complex applications areas is also outlined.