Title :
Task selection for control of active vision systems
Author :
Iwatani, Yasushi
Author_Institution :
Dept. of Syst. Inf. Sci., Tohoku Univ., Sendai, Japan
Abstract :
This paper discusses the task selection problem for active vision systems, that is, what tasks should be selected to control active vision systems and which order of priority should be set for the selected tasks. One possible task to determine an optimal camera placement is to obtain high resolvability, or equivalently, to optimize a perceptibility measure which is a quantitative scaling measure from error of the measured velocity of image features in the image plane to error of the corresponding velocity computed in the camera coordinate frame. This paper first shows that optimization of the perceptibility measure may produce unreasonable motion responses for active vision systems, and it should not be selected for a primary task to control active vision systems. This paper then proposes target tracking as a primary task and optimization of a perceptibility measure as a secondary task. The perceptibility measure for the secondary task is induced by certain Jacobian matrices, not the image Jacobian matrix, to produce cooperative behavior for active vision systems with multiple cameras.
Keywords :
Jacobian matrices; cameras; end effectors; image motion analysis; robot vision; target tracking; Jacobian matrices; active vision systems; optimal camera placement; perceptibility measure; target tracking; task selection problem; vision control; Cameras; Control systems; Coordinate measuring machines; Image resolution; Jacobian matrices; Machine vision; Motion control; Motion measurement; Target tracking; Velocity measurement;
Conference_Titel :
Intelligent Robots and Systems, 2009. IROS 2009. IEEE/RSJ International Conference on
Conference_Location :
St. Louis, MO
Print_ISBN :
978-1-4244-3803-7
Electronic_ISBN :
978-1-4244-3804-4
DOI :
10.1109/IROS.2009.5354150