Detecting Specified Object from a Cluttered Data in a Live Video Using Rotation Invariant Detector and Descriptor

Abstract

— This paper presents the detection of specified object from the cluttered data in a live video feed using Speed Up Robust Feature. In this method, for detecting an object point to point matching is done of the object to be detected with the cluttered scene. Computation and comparison made through SURF is much faster as compared to other algorithms proposed earlier. This is done by convolution of integral images, by using hessian matrix based measure for the detector and a distribution based descriptor and by simplifying these methods to perform the essential task. This concludes to new steps of detection, description and matching. This procedure is verified using a Logitech HD camera with Data Acquisition, Image Processing and Computer Vision toolbox in MATLAB(Trial Version). The experimental results show that this method works best for the objects that exhibits non-repeating texture patterns.