Method and system of identifying non-distinctive images/objects in a digital video and tracking such images/objects using temporal and spatial queues

摘要

A method and system to identify and locate images/objects which may be characterized as non-distinctive or “featureless” and which would be difficult to locate by conventional means comprises a plurality of steps including identifying first and second frame markers, increasing granularity between the frame markers, identifying at least one dominant object between the frame markers, normalizing its shape and identifying its edges, dissecting the dominant object into at least two equally sized sections; identifying shape and characteristics of at least one section (the analyzed section) of the dominant object thereby creating section data; applying geometric modeling such that section data from the analyzed section is used to determine overall shape, facets and configuration of the dominant object, thereby forming a geometric model; comparing geometric model to a known reference database of objects like the non-distinctive object (the reference object); and assessing the probability that the geometric model so formed represents the desired non-distinctive object.