Tagging and Tracking in Video Using Neural Net Color Detection and Spatial Filters

Tracking of objects in video is of interest in such applications as surveillance and automated machine vision environments. We have developed an algorithm capable of tracking an object in video. A neural net is generated from the first frame of a video sequence to detect the object of a chosen color. The object is then detected in subsequent frames using net weights determined from the first frame. The object is tagged and tracked using bounding spatial filters in polar coordinates.