Features Camera Corner Opinion
After analytics

Currently, the core value proposition of video analytics is being able to differentiate people and vehicles from other movement in a scene. Features like counting, intrusion detection, line crossing, etc. are all driven by this core functionality. The reality, however, is that there are technologies that already exist for identifying people and vehicles which are far better than video analytics.

The only challenge preventing wider adoption of alternative technologies is cost. Eventually the cost of alternative detection products will come down and quickly reduce the value of video analytics. With that said, video analytics has enormous long-term potential, if manufacturers are able to focus on the right features.

Assuming that in the future video analytics will not be the best way to detect people and vehicles, analytics providers will need to focus on services that require images or video footage to function. One good example is facial recognition. While there are a variety of facial recognition tools on the market, video (both visual and thermal) still offer long-term competitive advantages.

The most recent alternative to video based facial recognition is 3D recognition, which uses reflective light sensors. The challenge with this technology is that it requires the light sensors to be deployed at any location where facial recognition is required. This limitation is where video-based facial recognition will have a long-term advantage. While still in its infancy, the ability to run facial recognition software on pre-recorded video will ultimately make video based facial recognition through the use of video analytics the most efficient technology on the market.

Another value driver analytics will have over the long term is the ability to drastically improve the speed at which we search through recorded video. Before smart search tools existed, the only way to locate an incident in recorded footage was to fast-forward through the recording until you saw the activity you were searching for. Today, tools like motion and pixel search allow for much quicker searching but still have limitations. Even if a user is able to locate a suspect and an incident occurring on one camera, they are not easily able to track that suspect as they travel across other cameras on the same site. Further to that, if the police want to search for occurrences of that suspect across different surveillance systems on other buildings, the only way to accomplish this would be through manual video review. Video analytics is able to process video faster than humans and can drastically reduce search times by automatically searching for people with a similar appearance across multiple cameras or locations. Searching video is an overlooked challenge for many manufacturers, but for end-users and integrators who struggle through hours of recorded video in search of data, analytics could provide significant value.

Analytics may have gained recognition for its ability to differentiate people and vehicles from other motion, but in the long term this capability will become far less valuable as other technologies with better accuracy come down in price. If analytics providers focus instead on the areas where better data from existing video is required, they will find long-term success. The ability to map the activity of one person across multiple cameras or sites will always be valuable as video surveillance deployments continue to grow. Similarly, having the ability to search for an individual face in thousands of hours of footage will further improve the quality of data users are able to extract. Video analytics will soon face challenges from lower cost methods of detecting people, but analytics providers who are able to focus on the extended uses for video analytics like facial recognition and data filtering will continue to find success in an increasingly competitive market.  

Colin Bodbyl is the chief technology officer for UCIT Online (www.ucitonline.com).