A GPU-Based Statistical Framework for Moving Object Segmentation: Implementation, Analysis and Applications

被引：1

作者：

Cuzzocrea, Alfredo ^{[1
,2
]}

Mumolo, Enzo ^{[1
]}

Moro, Alessandro ^{[3
]}

Umeda, Kazunori ^{[3
]}

机构：

[1] Univ Trieste, DIA Dept, Trieste, Italy

[2] CNR, ICAR, Arcavacata Di Rende, Italy

[3] Chuo Univ, Tokyo 112, Japan

来源：

INTERNET AND DISTRIBUTED COMPUTING SYSTEMS, IDCS 2015 | 2015年 / 9258卷

关键词：

D O I：

10.1007/978-3-319-23237-9_19

中图分类号：

TP [自动化技术、计算机技术];

学科分类号：

0812 ;

摘要：

This paper describes a real-time implementation of a recently proposed background maintenance algorithm and reports the relative performances. Experimental results on dynamic scenes taken from a fixed camera show that the proposed parallel algorithm produces background images with an improved quality with respect to classical pixel-wise algorithms, obtaining a speedup of more than 35 times compared to CPU implementation. It is worth noting that we used both the GeForce 9 series (actually a 9800 GPU) available from the year 2008 and the GeForce 200 series (actually a 295 GPU) available from the year 2009. Finally, we show that this parallel implementation allows us to use it in real-time moving object detection application.

引用

页码：209 / 220

页数：12