Video Content Analytics: What it is and How it Works

Video content analytics (VCA) systems use deep learning to make sense of the potentially rich data stored in the reams of footage that often go unreviewed. The technology processes and analyzes the video to detect all the people and objects that appear, identify and classify them; and then indexes the video metadata so that it can be searched and analyzed. This is crucial for filtering out irrelevant details and pinpointing objects and people of interest. In short, this capability empowers users to locate key video evidence spanning many cameras in minutes vs. the hours otherwise needed to manually review each video feed. In addition, the technology can visualize the metadata as dashboards and heatmap overlays to optimize analysis, as well as support the configuration of real-time, rule-based alerts to trigger notifications based on predefined parameters so that store owners can remediate incidents in real time. 

Video Analytics in Retail

There are a number of ways that video analytics functionality can help retailers with their security operations. The technology can be used for people counting across a store and configured to trigger alerts when predefined traffic thresholds are exceeded. It’s an important capability, especially for tracking occupancy and proactively preventing overcrowding resulting from doorbuster sales.

VCAs can also trigger rule-based alerts when customers stray into areas that should be off-limits, such as shipping/receiving or the stock room. This can help security teams spot suspicious behavior during or outside of normal store hours and respond appropriately.

Dwell maps and heat maps can also help in several ways. They can differentiate between desired behaviors —  a shopper pausing to check out a special or popular item — or abnormally long times that might imply malicious intent or even a medical emergency. When combined with real-time alerts, these insights enhance situational awareness and aid intervention.

VCA can also accelerate investigations by uncovering details buried in the video. Sophisticated searching and filtering functions make it easy to scan hours of video in minutes by zeroing in on the right segments based on relevant criteria, including appearance similarity, clothing, color, size, speed, path, direction and dwell time, providing an ever-increasing and powerful set of distinct search combinations. Furthermore, a rapid forensic review can help gather evidence post-incident, saving time and boosting accuracy.

VCA in Action

When convenience store staff noticed a bottle of wine missing, they used video content analytics to isolate that shelf in their surveillance footage and rapidly search through hours of footage.

….where they found this woman putting the wine in her bag and walking out the door, for example.

Alerts can be triggered when a visitor enters an off-limits area….

…or when a stranger sneaks into the perimeter (see the man in the white shirt at 15.22).

This dwell map shows where people spend the most time looking at products. A sudden mass gathering in another section of the store might be cause for concern.

 Finally, this path map shows how customers move throughout the store. Based on how customers navigate the floor, store owners can optimize the space to prevent overcrowding and safety concerns. 

VCA  is a powerful technology that is helping businesses and organizations worldwide prevent theft, investigate security incidents and keep both customers and staff safe and healthy. The use cases for the retail industry are nearly limitless.

This article originally ran in Security, a twice-monthly security-focused eNewsletter for security end users, brought to you by Security Magazine. Subscribe here.