Object Recognition, Tracking and Trajectory Generation in Real-Time Video Sequence

Abstract

The aim of tracking is to detect objects through images. This paper focuses on detection of moving objects in a scene and tracking of the objects as long as they remain in camera view. Initially noise removal and image enhancement is carried out. The objects are detected from the supervised Euclidean segmentation of the image frames. An object recognition algorithm then classifies the segmented objects by employing minimum distance classifiers approach, and by comparing various length and shape descriptors. Finally, object trajectory is generated based on centroid location of geometric object. For the case study, we have considered tracking of a tennis ball and shown its trajectory after tracking. This method assures accurate image segmentation via specifying color intensities of the object. Although median based approach gives much better results but it's rather much more expensive. Same case is with Census Transform method that is computationally expensive and is complex if multiple objects are there per frame.