Deep learning has been shown to be efficient for multiple object tracking, despite the challenges of frequently occurring occlusions, uncertain appearances, objects in as well as out, and insufficient labeled data. Detecting and tracking objects is one of the most common and difficult jobs that surveillance systems must undertake in order to recognize important events and suspicious conduct, as well as automatically remark and extract video information. The progress of convolutional neural networks (CNN) changes the way objects are tracked. CNN layers trained upon a significant amount of videos or image sequences improve object tracking accuracy in shorter time periods. This study analyses and compares the network model and tracking techniques with its performance measures.

Multiple Object Tracking (MOT) Convolutional Neural Networks (CNN) Tracking Machine Vision Video Surveillance