An Integrated Framework for the Timely Detection of Petty Crimes

被引：6

作者：

Dimitriou, Nikolaos ^{[1
]}

Kioumourtzis, George ^{[2
]}

Sideris, Anargyros ^{[3
]}

Stavropoulos, Georgios ^{[4
]}

Taka, Evdoxia ^{[1
]}

Zotos, Nikolaos ^{[3
]}

Leventakis, George ^{[2
]}

Tzovaras, Dimitrios ^{[1
]}

机构：

[1] Inst Informat Technol, Thessaloniki, Greece

[2] Ctr Secur Studies, Athens, Greece

[3] Future Intelligence, London, England

[4] Univ Patras, Patras, Greece

来源：

2017 EUROPEAN INTELLIGENCE AND SECURITY INFORMATICS CONFERENCE (EISIC) | 2017年

关键词：

D O I：

10.1109/EISIC.2017.13

中图分类号：

TP [自动化技术、计算机技术];

学科分类号：

0812 ;

摘要：

While petty crimes are considered misdemeanors from a judicial point of view and are typically punished with light sentences, they greatly affect citizens' perception of safety and are related to substantial financial losses. In this paper, we describe a technological solution for the timely detection of petty crimes, based on the developments of the EU project P-REACT. Concretely, a modular framework is presented where an embedded system processes in-situ a camera stream for the real-time detection of petty criminality incidents and the timely notification of authorities. This paper provides details on the various hardware options and the key software components of the system, which include a set of appropriately implemented video analytics algorithms for the detection of different petty crimes as well as modules for the capturing, transcoding and secure transmission of video clips in case of an alarm. An evaluation of the system is also provided covering both experimental results on the accuracy of the platform but also focusing on the feedback received during the trials phase of P-REACT through the participation of external stakeholders. Evaluation during this phase was based on the live demonstration of system's operation in a series of simulated events corresponding to different types of petty crimes. In both cases evaluation results were very promising, attesting to the high innovation potential of the platform.

引用

页码：24 / 31

页数：8

共 17 条

[1]

[Anonymous], 2005, PROC CVPR IEEE

[2]

[Anonymous], ACM INT C MULT

[3]

[Anonymous], 2000, Pyramidal implementation of the lucas kanade feature tracker description of the algorithm

[4] Random forests [J].