| Abstract [eng] |
Video recognition systems are becoming more popular in everyday life. They are used in wide range of areas: traffic control, military, medicine etc. This paper is exemplified with the project “Universal Video Recognition Systems” developed by Liudas Drejeris, Ernestas Kardzys and Karolis Kriščiūnas. This paper concentrates on the problems encountered during creation of video recognition system: the selection of algorithm for video recognition, the selection of system architecture used, comparison of JSON and XML formats for information interchange. The created video recognition system is tested with three different kinds of objects (books, cups and toys) and is 92% accurate. System was created using Java and C++ programming languages. It uses SURF algorithm for image processing. System has clients for Windows, Linux and Android OS, server works on Linux OS. |
