This paper presents a new technique for tracking of two dimensional moving objects. The approach is a hybrid consisting of an algorithm for next position prediction using an estimated jerk, and a neural network for pose (position and orientation) determination. A three layer feed-forward perceptron with back-propagation is implemented as a mapping approximator to determine the object pose from the area information contained in a localized processing window. Objects with arbitrary shapes can be tracked after the network is properly trained. The experimental result of the implementation of this algorithm is also presented.
|Number of pages||4|
|Publication status||Published - 1 Dec 1992|
|Event||Proceedings of the 1992 Japan - USA Symposium on Flexible Automation Part 1 (of 2) - San Francisco, United States of America|
Duration: 13 Jul 1992 → 15 Jul 1992
|Conference||Proceedings of the 1992 Japan - USA Symposium on Flexible Automation Part 1 (of 2)|
|Country||United States of America|
|Period||13/07/92 → 15/07/92|