Video Character Recognition System

摘要

This report describes the development and testing of a video characterrecognition system (VCRS) designed to operate in a medium to high complex, non-linear noise environment. The video character recognition system used a feed forward, back propagation neural network to do the recognition. The noise generation allows high recognition rates even with limited example sets, significantly increasing the utility of the system. The intended use of the VCRS is to identify character fields within video data. Specifically, it is designed to perform data extraction for flight test and photo interpretation tasks. A sophisticated user interface has been implemented so that the VCRS can be quickly customized to a new character field or new video formats.