Abstract: This research paper presents a novel method for the identification of some object in a video using the distinctive invariant features from images. This method uses reliable matching between different views of an object or scene. The features shows a robust matching across a particular range of affine distortion, change in 3D viewpoint, addition of noise, invariant to image scale and rotation and change in illumination. In this, the recognition for the object proceeds by matching individual features to a database of features from known objects using a technique called as scale invariant feature transform. This approach to recognition can robustly identify objects among clutter and occlusion while achieving near real-time performance.

Keywords: Difference of Gaussian, Keypoint descriptor, Object, SIFT.