Food for thought: tips to accurate VCA

Food for thought: tips to accurate VCA

Many users purchase and employ video content analysis (VCA) in hopes that it would be the miracle solution to all their problems. As a result of the over exaggeration of VCA capabilities in its earlier days, users believed the technology will soon be able to replace human operators. They soon came to realize these claims were mostly propaganda, which resulted in great disappointment over the effectiveness of employing VCA.

Nowadays, the technology is more advanced and mature compared to its previous prototypes. So why are users still having trouble with their VCA? The answer is quite simple. Users need to be educated about the technology and its capabilities, so there are no misconceptions about what the VCA can or cannot achieve. “In some respects, we are attempting to use computers and algorithms to replicate what the human eye and a substantial part of our brains are used to accomplish. Certainly systems benefit from the automation achieved by today's video analytics but it is not a replacement for human decision making,” said Larry Bowe, President of PureTech Systems.

As humans are naturally equipped with the ability to read and decipher body languages of other people and register events as they occur, we automatically have the ability to register the meaning behind it and whether it is something that can or cannot be ignored. Video intelligence is not as flexible in that sense.“Video analytics is good at detecting clearly definable simple activities. Video analytics is not good at inferring intent,” stated Mahesh Saptharishi, President & CTO of Video IQ.

Users must have a good grasp of the type of VCA functions they will need for their applications. “Analytics are a tool that humans use, and at least at this point, are not intended to replace the human who makes decisions based on the information the system provides. The success of an analytics program is to know your application, know your environment, and know exactly what information you want the system to deliver,” said Shahar Ze'evi, Senior Product Manager at American Dynamics (Tyco Security Products).

Terms and Conditions
When it comes to indoor and outdoor applications, there are vastly different requirements from the VCA. “In terms of environment, lower end solutions are best used in simple, stable scenes such as intrusion detection in indoor areas. More challenging environments such as perimeter with dense moving foliage and shadows cast by vehicles on roads will often prove too difficult for lower end solutions,” advised Andrew Eggington, Director of Ipsotek. Oftentimes, disappointments from customers arise because they employ what may be better suited for indoor applications for outdoor use. An outdoor system must use stabilization to compensate for wind and vibrations; it must have geo-registration capabilities to create accurate size filters that ignore the movement of animals, trash, trees, and other irrelevant motion; it must be ruggedized to withstand temperature extremes, weather, sand, and dust, according to John Romanowich, CEO of SightLogix.

Depending on the application, thermal cameras with VCA is recommended for outdoor applications as it will ignore nuisances such as stray headlights and reflections that would otherwise set off the alarms on visible cameras. Thermal cameras are also great for environments with little or no lighting, as it will be able to detect people and objects better than regular cameras in complete darkness.

For VCA to achieve optimal results, the cameras in use must have good picture quality and picture content. The VCA solution should only be deployed in places where they will generate value. Most of the time, employing analytics will result in a change in users' current camera deployments. “Customers should consider camera positioning if they are thinking of using analytics. Analytics needs a clear view of the target and many existing surveillance cameras are positioned too far away or the cameras are too low so that the targets are hiding each other,” said Geoff Thiel, CEO of VCA Technology. Take counting as an example, “counting can only be reliable if the camera is installed above the scene, ideally on top. Looking horizontally it will not give good results. The algorithm needs a certain time to analyze the scene: The longer the detection, the more precise the analytics,” said Achim Hauschke, CEO of Riva. If the camera can be easily blocked by other objects or people passing through, the VCA will be rendered useless.

Reasonable Expectations
Since VCA is not a plug-and-play technology, an experienced vendor will not only have robust technology, but also know how to plan and deploy successful VCA projects, according to Illy Gruber, Product Marketing Manager at Nice Systems. “Ask for a performance commitment from vendors, though it is tricky because vendors will not commit ‘blindly' to any performance, but experienced vendors should be able to indicate overall expected performance in the user's specific environment based on site survey and proper planning.”

Also, users need to understand that VCA, no matter how high end, is susceptible to false alarms. “A sensitive system will always produce a small number of false alarms, which can be minimized but not completely eliminated if the VCA is also to detect,” said Frank Brandtner, Head of Administrative Product Management at Geutebruck. There are different approaches users can take to decrease the percentage of false alarms in their settings. Some vendors will offer free evaluation periods to customers to best educate potential customers about their products while reviewing customers' requirements to offer the best recommendation, according to Sadiye Guler, Founder President of intuVision. Users can take advantage of this offer and test to see if the VCA will really be beneficial to their applications.

Once users are aware of the conditions to obtain the most accuracy in their applications, they will be removed of their former misconceptions and expectations of VCA features and functions.

Share to:
Comments ( 0 )