Here is the next part of the previous blog series, Thanks again for reading the blog…
Continuing with our previous discussion on the pioneering technological breakthrough achieved by developing Cognitech VideoActive, the session aims to explain the essential features of this Real-Time Video Forensics Software as wells as Crime Scene Reconstruction Software.
With the exponentially increased use of CCTV cameras across all functions and departments, it becomes difficult for the security personnel to monitor and ascertain the situation distributed through multiple cameras accurately. This is where Real-Time video analysis finds its natural use.
Patented Automatic Demultiplexing
Another application that is naturally suited for Real-Time implementation is Demultiplexing, which converts intertwined multiple camera video source streams, such as multiplexed CCTV footage, into individual digital camera channels that can be stored, processed or viewed, individually or together. The multiplexed videos are still to be found with the analog recording system. Automatic demultiplexing intelligently determine how to separate demultiplexed frames into individual camera channels with no human interaction for any number of channels. Manual demultiplexing, on the other hand, sorts demultiplexed video according to user-defined key-frames.
The first-ever scientific publication on Demultiplexing and specifically Automatic Demultiplexing through algorithm was published by Cognitech’s scientists: “Software-based universal demultiplexing: threshold-free energy minimization approach”
Author(s): Frederic Guichard; Alexander Litz; Lenny I. Rudin; Ping Yu
2001, SPIE Proceedings Vol. 4232 Conference on
Enabling Technologies for Law Enforcement and Security,
The original idea of Automatic Demultiplexing was to compare the streaming video frames and group them accordingly to similarity criteria. These similarity criteria can be as simple as correlation measure between frames, which assumes that frames do mot have dynamic changes, like moving vehicles or walking people.
Cognitech scientists improved the demultiplexing algorithm further to account for scenes with constant changes present, like casino floor and freeway traffic (published in Advanced Video and Signal Based Surveillance, 2005. AVSS 2005. IEEE Conference
“Meaningful automatic video demultiplexing with unknown number of cameras, contrast changes, and motion.”Jose Luis Lisani, Lenny Rudin, Pascal Monasse, Jean-Michel Morel, Ping Yu)
United States Patent Office has awarded Cognitech US Patent #820059 L. Rudin et. Al.
“Video demultiplexing based on meaningful modes extraction” for an algorithm that is the basis of the Cognitech VideoActive Demultiplexing module, the most robust proprietary Demultiplexing software on the market.
Fast forward to 2019. Cognitech scientists have been extending the earlier methods for Automatic Demultiplexing to such important tasks as Video Search, and Sort and Indexing.
If the camera is PTZ, it is necessary to be able to group frames that do not even have the same content, but ‘linked’ spatially:
Once one or several of the video are Demultiplexed, the end-user can view these sorted camera feeds through Cognitech Video Investigator’s VSI Interface. VSI is a Video Scene Integrator, a specialized Video Forensics Software that is organizing CCTV to facilitate the investigation process of videos from multiple cameras, usually adjacent in space and time, like tracking the suspects that enters the department store and walks across areas covered by different cameras.
Manual tools are eventually to be supplemented by automatic Tracking and Video Search Tools. US Patent Office has awarded Cognitech a very versatile US Patent to Cognitech: L. Rudin et. Al.” System and method for image and video search, indexing and object classification” US Patent number: 8831357
These are also naturally Real-Time applications. For example, one may need to detect in the hours-long video each individual using an automated teller machine, assign this person an index, and store all indexed individuals into a database for subsequent retrieving based on attributes such as human height ( with automatic crime scene measurement tools). Similarly, the passing by vehicles can be detected, indexed, and stored. Moreover, identifying attributes can also be extracted, such as vehicle color, dimensions and eventually Make and Model, a good application of video forensics analysis software.
We hope you enjoyed this CogniKnowledge Blog! Stay tuned and don’t miss the next blogs. You can also follow us on Twitter, Facebook, Instagram, LinkedIn, or YouTube: we post CogniKnowledge Blogs and ForensicFriday Blogs regularly so you won’t miss any!