Cognika was founded in 2006 with a mission to bring leading edge analytics tools to market that had been developed at MIT. Our breakthrough technology uses Artificial Intelligence (AI) techniques with our patented algorithms that emulate human cognition and serve as the foundation for the Pegasus/Perseus software tools.
Cognika has significant experience in building technologies around managing "big data" and analysis. We are focused on exploitation of Video & Imagery, and do so through a combination of "real-time" monitoring and forensic look-back ("search") capability within video archives. Our technology scales well with customer needs to Petascale and larger volumes.
Pegasus is our video management tool for moving camera platforms (e.g. aerial) and Perseus for persistent surveillance (e.g. surveillance cameras) . Both tools provision visual search and discovery with real-time alerting for activities of interest e.g. "Persons Gathering", "Vehicle making U-turn". Complex activities (collections of events) are also detected in real-time.Learn More
Cognika makes video & imagery exploitable in exciting new ways. We make video "searchable" and analyze it in real-time to detect & alert for marked objects, events and their interactions: "Activities". For example, we detect if a white pick-up truck has repeatedly appeared around an area of interest, or if there is anomalous activity within a certain region from a wide-area asset, or forensically track-back a vehicle of interest across multiple cameras and videos. More complex activities are also monitored and indexed.
With millions of minutes of video being generated from an increasing number of sensor platforms, the rapid increase in video is overwhelming human analysts, necessitating automation and augmentation tools. Our approach utilizes a novel machine-vision based approach to index Full Motion Video (FMV) which provides the user the ability to detect, search for and track objects of interest, events and activities . This approach enables FMV & imagery exploitation in real-time, with a forensic look-back. Available metadata provisions advanced query capabilities such as Geo-filtering, time-filtering to zero-in on Events/Objects.Learn More
Cognika Product Family
Pegasus is designed for dynamic camera feeds such as from aerial assets, vehicle mounted etc. It applies sophisticated stabilization, mosaicing and tracking algorithms to detect objects, events and activities. Pegasus also can alert when a specific Object of Interest (OOI) appears in the Field of View.
Pegasus applies Cognika's patent pending "Visual" search and monitoring technologies to locate when a specific type of vehicle or person appears. Furthermore, it instantly searches video archives (from multiple camera feeds) to locate similar OOIs providing a time-line for analysis. The design elegantly bridges real-time and forensic analysis and empowers users to exploit video and imagery in conjunction with Multi-INT sources. Pegasus also offers RESTful interfaces for easy integration into existing applications. Pegasus is available for Windows and Linux platforms.
Real-time Object Detection within Video. Demonstrates the powerful capability of locating OOIs even if they are in FOV for a few fleeting seconds.
Works with FLIR video, of low quality. Even with "blob" like quality and works with assets such as UAVs or similar assets..
Tracking a rapidly moving object from an aerial asset. Works despite Occlusion and different poses. Works with move-stop-move scenarios common in urban environments.
Perseus is designed for fixed cameras performing persistent surveillance. Potential applications include Perimeter Security, Area Surveillance, Building or infrastructure security. Next Generation of Video Analytics tools for Activity Recognition and Object Search.
Perseus is designed to work in both real-time and forensic modes simultaneously. It eempowers analysts to view an OOI or Event in conjunction to view where/when else the OOI was possibly spotted. This amalgam of augmented analysis is at the core of Cognika's model of breaking down the traditionally silo-ed approach. Perseus offers RESTful interfaces for easy integration into existing applications. Perseus is available for Windows and Linux platforms
Cognika Makes information findable in a ?Google-like? fashion in (near) Real-time. It is Multi-modal: processing textual (documents, emails, IM etc.), Video, and structured data, Multi-INT: Fuses information across HUMINT COMINT, SIGINT data etc., and Ad Hoc: provisions efficient cross-network, cross-domain information flow via search and discovery while dynamically provisioning for appropriate security restrictions and audit-trail.
The powerful technology provides a system for very sophisticated analytics in order to make reasoning/inferencing more intuitive, accessible, manageable and meaningful.
The Office of the Undersecretary of Defense for Intelligence, OUSD (I), defines "Activity-based Intelligence" (ABI) as "a discipline of intelligence where the analysis and subsequent collection is focused on the activity and transactions associated with an entity, population or area of interest.? CIDS' approach provisions recognition of sequences of events within video such as, for example ?vehicle u-turns?, ?persons gathering? etc. This activity is augmented with multi-INT data relating to the activity at hand in a dashboard fashion providing the analyst a complete picture of the situation.
Cognika's Perseus suite is capable of monitoring video streams from from both aerial and fixed camera assets. We employ sophisticated algorithms such as stabilization,enhancement,OOI identification etc. to detect patterns-of-life and correlate information across multiple assets.
We support both EO & IR format video and our algorithms work from HD quality to low resolution gray-scale video. For very noisy video, we apply pre-processing techniques to enhance video quality. We process both archived (stored) video and/or streaming video with most major video formats (H.264, AVI, MPEG, Quicktime, WMV etc.)
Cognika's Perseus suite is capable of monitoring video streams from a certain region detecting events that might be of interest and alert based on triggers. Furthermore, it can retrieve relevant information from multiple sources dynamically to build a "Virtual Profile" of a person, vehicle or event. Exemplars of activities of interest include: anomalous human movement, abnormal traffic on a road etc.
We offer both SOAP and REST-ful interfaces into our application server. Our solution is available either as an appliance or an installable suite for bespoke application development, and we offer standard GUIs as part of our stand-alone offering.
Our technology is based on best-practises proven in products such as Netflix, Twitter, Facebook etc. We use a combination of proprietary and open-source tools for our real-time processing and scale to Petascale (and larger) data-sets. Even on such scales our query times are typically in the order of few milliseconds with very tiny latencies.
Cognika applications allow knowledge workers to make sense of complex information in ways that traditional interfaces can't. Data that may seem too overwhelming, too multi-faceted, or too complex. Cognika makes information more intuitive and understandable, often leading to discoveries of relationships and nuance otherwise infeasible.
Data integrity across MULTI-INT sources involves dealing with inherent uncertainties both in interpretation and communication. While existing approaches such as Ontologies provide shared representations of the entities and relationships characterizing a domain, into which vocabularies of legacy systems can be mapped. However, a major limitation of traditional ontology formalisms is the lack of consistent support for uncertainty. Because uncertainty is a fundamental aspect of multi-INT fusion and a de facto condition under which such systems operate - this is a serious deficiency. Current ontology formalisms provide no principled means to ensure semantic consistency with respect to issues of uncertainty or data quality. Ontologies provide the ?semantic glue? to enable knowledge sharing among diverse systems cooperating in data rich domains such as Intelligence Analysis, but fail to provide adequate support for uncertainty, and ubiquitous characteristic of open world environments. Effective multi-INT fusion requires uncertainty management to be effective, and our approach provisions such an effort.
As the phrase indicates, the ESP platform is capable of gleaning hidden or hitherto unknown relationships between concepts using machine-learning and statistical methods. These relationships can be verified by a human analyst by examining the chain of relationships & predicates the system uses to achieve its inference. For example, a scientist may learn which genes have the highest correlation with a certain medical condition, which biomarkers have shown most promise in relation to detecting a chronic disease. Since knowledge is dynamic and evolving and the sheer scale of it overwhelming. For example, there were almost 700,000 articles published in PubMed in 2008 alone, it has therefore become impossible for researchers and scientists to keep up with the sheer volume of information. Cognika's ESP platform makes this information deluge consumable.
Quick introduction to Perseus and its exploitation capabilities. Offers perspectives on multiple use-cases for Perseus.Create timeline of events and build a storyboard for forensic analysis post fact. A powerful combination of real-time and forensic search.
Detects human and vehicular activities in a parking lot. Useful for learning "patterns-of-life" and alerting to anomalies. Sophisticated alerting capabilities can alert on detecting a specific object of interest (e.g. white pick-up truck, person carrying box etc.).
Persistent surveillance with video, of low quality. Even with "blob" like quality and works with assets such as UAVs or similar assets. Provide "stare" capability to compounds or areas of interest.