Identifying and Addressing Challenges for Search and Analysis of Disparate Surveillance Video Archives

Abstract

This paper discusses the challenges faced when bringing together multiple disparate surveillance video archives to support semantic analysis and search and describes the SAVASA framework for enabling better integration of CCTV archives. The proliferation of CCTV cameras managed by public institutions and private enterprise raises a number of issues relating to data security, privacy, ethics and technological difficulties in unifying the variety of data and formats. These are often the result of misunderstandings between the involved parties and include concerns such as reliability and security of cloud technology, accuracy and privacy in automatic semantic annotation of video and ultimate responsibility over content in the digital age. In this paper we present outcomes from the SAVASA project that aims to develop a standards-based video archive search platform allowing authorised users to query over remote and non-interoperable video archives of CCTV footage from geographically diverse locations. Accessing disparate surveillance video archivesThe increasing ubiquity of CCTV and surveillance video systems results in very large archives of footage captured and recorded in remote locations, at different levels of coverage and with different formats, available metadata or searchable indices. Authorised users face many challenges accessing specific footage or finding relevant segments based on semantic descriptions such as "white car", "person running". Both research and commercial efforts have sought to improve automatic analysis, management and information extraction from digital video footage [10,8].The SAVASA 1 project -Standards-based Approach to Video Archive Search and Analysis -aims to develop a standards-based video archive search platform that allows authorised users to query over various remote and noninteroperable video archives of CCTV footage from geographically diverse locations. At the core of the search interface is the 1 http://savasa.eu application of algorithms for person/object detection and tracking, activity detection and scenario recognition. The project also includes research into interoperable standards for surveillance video, discussion of the legal, ethical and privacy issues and how to effectively leverage cloud computing infrastructures in these applications. Project partners come from a number of different European countries and include technical and research institutions as well as end user, security and legal partners.Recent developments in CCTV systems have resulted in higher resolutions and new technologies to can take advantage of analytical algorithms that monitor footage and provide alarms or automated actions through previously setup triggers. Furthermore, CCTV devices are becoming Edge devices (entry points), storage systems and/or off-the-shelf servers instead of proprietary DVR (as in the past). This means video processing can be fully integrated and supervised by software systems.This paper looks at the challenges of providing integrated search of surveillance video ...

Retrieval in Long-Surveillance Videos Using User-Described Motion and Object Attributes

“…Although these methods could probably be adapted to index the video and facilitate search, very few methods this problem explicitly [18]. An exception is a method based on topic modeling by Wang et al [32].…”

A Real Time Visual Exploratory Search Engine for Information Retrieval in a Cloud

“…The two components share a store component, which stores all data: document contents, and metadata instantiations, and intermediate results created by the analysis and annotation components. Little [17] recommended in his study the use of Semantic Framework and semantic annotations. This study was more relevant to information retrieval among Videos and this study proposes to adopt the same concept along with the study of Tang [4] which had certain limitations.…”

Automatische gedragsanalyse voor effectiever cameratoezicht in de openbare ruimte

“…Tenslotte is de mate van interoperabiliteit een factor van belang bij samengestelde systemen. Als een toezichtsysteem bestaat uit verschillende subsystemen dan moet extra aandacht worden besteed aan de koppelvlakken om de beelden geschikt te houden voor VCA (Little et al, 2013).…”