Efficient, automated video surveillance
An EU team created a novel system of linked encoders and analysers, optimised for automated video monitoring applications. Outcomes included efficient new pipeline software offering various performance improvements, such as superior tracking and stabilisation.
Several technical issues affect the efficiency and costs of video
surveillance systems. Compression systems are generally separate from
analytical elements such as face-recognition, while also being optimised
for human viewing rather than machine analysis.
The EU-funded SMARTENC (Smart Video Encoders for Wireless Surveillance Networks) project aimed to rectify the situation. The consortium planned to design video encoders and analytics systems, both intimately linked with surveillance cameras. The designs were intended to improve performance of video analysis in various engines, and compression quality.
Work involved development of new encoding techniques and streaming algorithms. Results included pipeline software called Object Oriented Video Processing Architecture (OVA). The software handles video acquisition, processing, analysis and display. OVA was also incorporated into two other EU projects (ITEA2 and SPY).
As components of OVA, the project designed and developed four algorithms for real-time video analysis. The innovations allow detection and tracking of moving objects, video stabilisation, and generation of panoramic video. An additional algorithm, developed in collaboration with a commercial organisation, improves efficiency of real-time video encoding.
The project also devised a camera system, Smart Open Networked Camera, which will enable field-testing the software. Work resulted in several patent applications being filed, and conference publications.
SMARTENC yielded novel systems and products addressing the inefficiencies of conventional video compression. As a result, automated video processing for surveillance will be cheaper and more efficient, potentially leading to greater public safety.
last modification: 2016-04-22 14:07:17