Adaptive threshold based segmentation for video object tracking
Deepak Gambhir, Meenu Manchanda · 2014
An automatic segmentation and color feature based video object tracking algorithm has been proposed. The proposed algorithm automatically segments the moving object in video by creating a multiplicative mask, which contains reduced number of shadowed pixels, noisy pixels and false pixels. The segmented object can be tracked by extracting its features such as color. Once the object to be tracked is segmented and its feature extracted, the position of the moving object is predicted using Kalman filter which is an optimal recursive estimator. Kalman Filter efficiently tracks the moving object in real time applications. The proposed algorithm accurately segments the moving object by reducing the effect of the shadowing and/or noisy pixels and successfully tracks the moving object.