Classification of Moving Objects Using Recurrent Motion Image Classifier

In advanced security and automated surveillance systems, Object Recognition and Classification plays a vital role. Motion Based Recognition has been improved by using Recurrent Motion Image (RMI) classifier. For efficient Object Recognition the input video frames endures Preprocessing method such as (i) Background Subtraction using Linf distant image method, (ii) Shadow Points located by transforming the pixels from RGB (Red Green Blue) color space to HSV (Hue Saturation Value) color space and these points are remove by exploiting Gaussian Filters, (iii) Foreground Blob extraction by incorporating Connected Components Labeling algorithm and Blob Analysis using Blob size threshold to filter noise clutters. Postprocessing method encompasses (i) Blob Tracking by Region Correspondence algorithm, (ii) Blob Classification using RMI based on their periodic motion patterns. Most of the existing classification algorithms deal the Object Classification with the constraint of classifying only Human Being, Vehicles. But, the proposed algorithm has been extended such that it violates this constraint by classifying FourLegged Animals in addition to Human Being, Vehicles and upgrades the Object Classification methodology. By implementing this proposed algorithm, all moving objects were classified into proper categories as (i) Single Person, (ii) Group of Persons, (iii) Vehicles, (iv) FourLegged Animals.