Automatic Face and Object Identification and Motion Tracking for Indexing Digital Video

点击翻译按钮获取中文摘要

Although video, television and film have developed from an analogue format into a digital viewing experience, the medium itself remains largely incomprehensible to machines.Whilst there have been huge advances in the ability to produce, distribute, store and display digital video cheaply and effectively, there is almost no intelligent digital understanding about the visual analogue objects video contains. The moving image that this data represents can only be fully understood by humans and at present video content is searched for and categorised by information contained within a title or filename and an often illusive description. The traditional medium doesn't have the characteristics of other data-rich assets we find in today's modern digital environment.Adding a digital stamp or reference point to moving video objects creates a link between the visual, two-dimensional, analogue shape and the power of third party digital data. This project aims to develop a tool that allows content owners and producers to analyse and process video files, from production rushes to final composed edits, so that picture information can be stored, moving people and objects can be; identified and recognised, their movement recorded and a digital reference point added. This allows content to be indexed and searched for by picture, people and objects it contains.