Efficient Object Recognition Using Parametric Eigenspace under Influence of Noise and Occlusion

The effect of noise and occlusion on parametric eigenspace is studied in depth in this paper in order to identify the response of the system for distorted inputs from different sources and environmental factors. The eigenspace method of Murase become very popular due to its ability of automatic visual learning and power to capture the parametric details of the object with low memory. The work done is aimed towards analyzing the different factors that may affect the performance of the system. Paper discusses the effect of number of images used and effect of number of principal components used while training, on recognition system with highly correlated objects with experimental analysis. The work is also carried out for the effect of variation in universal PC for noise and occlusion with correlated and non correlated objects. The work also focuses on memory utilization of system.