A new support tool using object tracking and motion based segmentation is developed for machine learning
and pattern recognition. In the learning step, an object of interest is tracked while learning is performed from
segmented frames. In the recognition step, target is tracked until favorable conditions allow identification. This
tool is used in the context of the Aqu@thèque project which includes an automatic fish recognition system.
Tracking is a dificult task especially in case of real world images. Particle filtering methods incorporating
motion based segmentation measurement in importance sampling step improve performance.
In this paper, we suggest a new importance sampling scheme to improve a particle filtering based tracking process. This scheme relies on exploitation of motion segmentation. More precisely, we propagate hypotheses from particle filtering to blobs of similar motion to target. Hence, search is driven toward regions of interest in the state space and prediction is more accurate. We also propose to exploit segmentation to update target model. Once the moving target has been identified, a representative model is learnt from its spatial support. We refer to this model in the correction step of the tracking process. The importance sampling scheme and the strategy to update target model improve the performance of particle filtering in complex situations of occlusions compared to a simple Bootstrap approach as shown by our experiments on real fish tank sequences.