Recognizing Objects by Detecting Multiple Moving Parts

Detecting objects in a given video stream is an important step to understand their types, movements, and activities. Existing object detection algorithms suffer from their inability to detect the components constituting a particular object that may result in classifying such components as standalone objects. Such instances may happen particularly when the colors of some components of the object have colors, which are close to the background. In this paper, we propose a technique to detect such objects by analyzing multiple images for the same object and observing the motion of various components of the