Improvised Technique for Accurate Back Ground Subtraction Using SVD

Abstract

Moving objects tracking in video sequence is a major problem in computer vision, with applications in several fields, for example video surveillance and target tracking. Most of the techniques published in the article use background subtraction techniques to obtain foreground objects, and they have put into use shadow detection algorithms and have analyzed spectral information of the images to fetched only valid moving objects. In this paper, we have come up with a small improvement for the current background model, and integrate a novel technique for shadow detection in grayscale video sequences. The suggested SVD (singular value decomposition), algorithm works well for both indoor and outdoorgarrangement, and it will not require the use of color cameras. We have tested and evaluated the performance of the suggested method, with standard threshold detection background subtractions algorithms. The preliminary results showed that the suggested method provides satisfactory output.