Public Funding for Ipsotek Limited

no public description

Ipsotek specialises in advanced video analytics solutions enhanced through artificial intelligence techniques ("AIVA"). It's technology has been applied to address safety, security and operations requirements in a wide range of commercial verticals, and particularly in airports where it has significant deployments. Ipsotek intends to apply its technology to provide bespoke solutions for the UK rail industry by developing a suite of capabilities that address immediate challenges and support long term growth plans. Safety and security challenges will be addressed with the application of Video Analytics (VA) and Artificial Intelligence (AI) technology to accurately detect and identify each threatening behaviour or safety risk demonstrated by individuals or crowds at the station. Here specific AI networks will be deployed that are trained to detect people and other objects in crowded scenes, estimate crowd density and generate density heatmaps. These networks will be further trained on Network Rail specific camera footage to improve their performance and make their experience specific to the rail CCTV environment. Furthermore, specific VA algorithms will be customised to raise events when behaviours of interest have been observed. Ipsotek will fully utilise its patented Scenario Based Rule Engine to create precise and detailed scenarios of interest that will address immediate requirements while providing the flexibility to adapt as the behaviours evolve and requirements progress. Additionally, on a continuous basis the system is able to generate data based on scene conditions that will enable crowding and other trends to be identified, as well as generate real time alarms for predefined safety and security threats detected. The developed solutions will be deployed and demonstrated in a live station environment on live feeds from the station CCTV network with live events and reports provided to station staff and managers. Throughout the project, Ipsotek will demonstrate the feasibility of adopting the solutions developed on a large scale with minimum reconfiguration or redesign by relying on its mature underlying video analytics platform that provides advance scene recognition, segmentation and calibration capabilities. Ipsotek aims to provide a system that will assist rail managers to reach informed decisions related to station and passenger management, enhance security and safety by heightening situational awareness and improve the overall rail journey of commuters.

Airport Suite

Ipsotek Ltd (Ipsotek) is an established market innovator with unique video content analysis techniques and IPR based on enriched video metadata. The key objectives/aims of the project are to be able to accurately, automatically and retrospectively ‘Tag and Track’ (TnT) specific moving objects through an area covered by multiple cameras. The TnT project aims to extend these concepts by developing new solutions for; 1) Scene understanding; 2) Crowded scene analysis; 3) Algorithms and models to enhance metadata and analysis performance; plus undertake 4) Site based operational demonstrators for validation of performance. The majority of present day intelligent ‘detect & alarm’ systems provide only rudimentary automated triggers. In other areas video used for retrospective event analysis and forensics requires close scrutiny of footage which can currently only be achieved manually. This is also limited to single channel (camera) content and is therefore time-consuming, costly and errorprone. The innovation and novel aspects of TnT will include: 1) Solving the problem of tracking objects in a single crowded scene. This is a limitation in all video analytics systems that significantly hampers deployment in real world environments. This alone merits consideration for advancing Video Analytics generally but when applied to the project it will significantly lift the commercial opportunity; 2) Unique and innovative metadata enrichment; 3) Increasing intra-camera recognition for effective enhancement of appearance models. The main benefits will include: 1) Provision of true capability for real time and retrospective automated moving object tracking with enhanced object identity data across multi-camera scenes with complex backgrounds; 2) Considerably more effective forensic video tracking capabilities saving time and increasing search success rates; 3) A robust framework on which to develop and productise practical applications having solved key technology roadblocks