It wasn’t so long ago that investigators thought themselves fortunate to find surveillance video relevant to their cases. Now the problem is more of what to do with all the available video information. Fortunately, video analytics systems have advanced to make this embarrassment of riches manageable.
Picking people out of a crowd
Deep learning technologies incorporated into video analytics systems will find details that previously required human eyes. These systems can distinguish men from women, children from adults, and sometimes even elders from younger people, based on size, gait, walk speed and other subtle factors people distinguish intuitively. On command, they filter people dressed in red from those in other clothing, or find just the pedestrians carrying backpacks.
This alone is a powerful resource. If an investigator is looking for a male suspect, dressed in white, and carrying a backpack, he might need to review hundreds of hours of video from multiple cameras. Analytics permit him to focus only on people who match the description, eliminating others from the feed.
Once someone has been identified as a person of interest, the system can locate that same person on other video feeds and segments, even if they span multiple days and locations. It becomes relatively easy to isolate the individual and trace his movements, so long as they appear within the feeds under analysis.
Identify high-crime locations
Finding sites where illegal narcotics sales are taking place has traditionally required a stakeout team to watch blocks and street corners for days at a time, hoping to document a series of apparent sales before sending in an agent to make a controlled buy. This process can now be largely automated by posting a covert surveillance camera and having the analytics system watch for telltale foot traffic. If there are frequent trips between a corner and a mailbox or nearly parked car – even if they’re made by multiple individuals – that location is probably where the drugs are concealed and retrieved from when the sale is made.
Conversely, foot traffic at a time or place when people tend to be scarce may indicate burglars casing a potential victim location. The analytics systems can even produce this data in real time, providing valuable and actionable intelligence to the street crime or patrol force.
As the baby boomers age, the incidence of seniors wandering away from their caregivers becomes more commonplace. A video analytics system can identify pedestrians who are moving unsteadily or slower than normal, and alert patrol officers to perform a welfare check, sometimes even before the caregivers know their charge is missing. Rapid response to these situations save lives, and are golden for public information officers always in search of good news to deliver to a critical public.
Urban planning and critical incident response
Estimation of crowd size and quantifying the flow of vehicle and foot traffic in an area is usually not much more than a “guesstimation.” A video analytics system provides defensible numbers to present to local government officials, and allows tactical planners to better allocate resources for public order situations. These analyses can be done in real time, or be based on archived video. This not only saves money, but provides better quality data.
Machine-based video analysis of the quality and quantity available today was not possible at all just a few years ago, no matter how much computing power one might have had available. They provide critical information for the people charged with the public safety.