Web searches tend to produce better results when the query is specific. For instance, submitting “mercury” into a search engine would bring up web pages referring to the chemical element, the planet, the car company, and more. Adding a second search term is obviously helpful; searching for “mercury rotation time” most likely wouldn’t bring up web pages regarding the car company or element. Searching through video footage works very similarly when the video is tagged with descriptive metadata.
Unlike traditional CCTV surveillance systems of the past, no longer do security personnel need to sift through hours of footage to locate a key incident. Video management solutions today employ advanced video analytic software that automates the cataloging of footage for more efficient searches later. Video analytics detect when video footage contains motion and assigns specific metadata—data about data—to the clips.
Analytics are intelligent enough to recognize that a large, smoothly moving box should be classified as a car and that a smaller, upright being with limbs would be deemed the label of a person. Depending on the quality of the camera and the processing abilities of the analytics, metadata could also include other information such as speed, actions, colors, locations, and even individual identities if integrated with facial recognition software.
The ability to perform an attribute-based search of an event such as “red car enters parking lot” out of a pool of 30 days of footage data and display the results within seconds is extremely helpful. However, video analytics are not yet fully autonomous or perfect; there will always be questionable situations and false alarms which require a human to remain in the loop of surveillance processes.
Related blog post: Focus on video analytics: License plate recognition