Detekce pohybu ve videosekvencích

Abstract

This paper describes the basic methods of motion detection in video sequences and discusses different approaches to solving this task. There are described advantages and disadvantages of different approaches and their suitability to specific examples. Furthermore there is described the article Abnormal Motion Detection in an occlusive Environment as a basis for the implementation and expansion of this method. Part of this work is the implementation of the analyzer as a cloud based service using NodeJS and native module in C++. This connection provides easy scalability and also enough computing power. There is also analyzed the appropriate approach to the development of such services and described the process of application assembly automation and configuration of infrastructure. This work also includes test for analyzer on several existing datasets, including the production of own one.

Description

Import 03/11/2016

Subject(s)

Optical Flow, Foreground and background separation, DBScan, K-means, C++ and NodeJS interoperability, Cloud, Atomation

Citation