The Big Data Hiding In Your Video Security Camera


By Eric Olson, Vice President Marketing,  PureTech Systems

Eric-242x300 The Big Data Hiding In Your Video Security CameraIn today’s environment video is everywhere.  In our offices, parking lots, supermarkets, roadways, neighborhoods, gas stations.  Typically, these sensors are deployed for security purposes, but their real potential as a rich supply of big data analytics is often overlooked.  A video stream is a gigantic pipeline of information, pushing huge amounts of data packets 30 times every second.  The key to extracting meaning from this pipeline of big data is through a technology called video analytics.   This technology processes video pixels and converts them into actionable data for all types of businesses.  Video analytics has the ability to discern different types of objects from a video stream and understand many of their actions.  This can be achieved in both real time and through analysis of archived video. Let’s take a look at six types of big data that your security cameras could be providing you.


Customer Demand Behavior

Collecting and analyzing simple customer behavior is by far the easiest and most popular data that can be derived from your video system.     Video can be used to count people, vehicles, or both.  In addition to just counting, heat maps showing preferred routes and areas of highest activity allow managers to analyze customer behavior and change processes or change the level of support resources.  The ability to understand location within the video (often referred to as location-based or geo-referenced video) allows for more precise data analysis, such as the ability to correlate purchases and customer routes – customers who purchased higher priced wine also spent more time in the fresh produce area.


Customer Classification

Personalized advertising or actions is a fantastic way for retailers to utilize big data.  In addition to providing valuable trend data, using video-derived big data to identify object classification, age categories, gender or number of customers in real time affords the ability to take immediate actions to increase sales.  For example, a fast food retailer with LED menu boards, can use real time video data indicating a large number of children has arrived and change the point of sales menu board to reflect choices more preferred by this age group.  Similarly, parking ramps can use video classification data (Car, Motorcycle, Truck) to understand trends relating to parking space size allocation.  The same real time data can trigger signs directing a certain vehicle type to the location in the garage with an accommodating space.


Point of Sale Conversion

Big data derived from video analytics algorithms lends itself to assessing new in-store advertising or procedural effectiveness.  A simple correlation between customers who visit a retail location end cap, pausing to observe, can be compared against actual sales data of that item to evaluate how well it is converting.  This can be further studied by including customer classification data, such as gender, age or even if they are utilizing a shopping cart.


Customer Service Performance

An important feature of most businesses is customer service.  Video-derived big data can play a role here as well.  In addition to simple tasks like determining queue wait time, intelligent video can also provide data to allow more advanced service analysis and facilitate preventative actions.   For example, by deriving location data from video, meaning the software understands the real location of the object in the video, customer actions such as loitering can be determined.   A user now has the data to know when a customer has been standing at an exact shelf location for a period of time.  This may indicate that they have a question, or perhaps they are up to no good.  The length of time they loiter, as well as, the response time for a customer service rep to respond, is extremely valuable trend data.


However, it does not have to stop at trend data.  This type data used in real time, can identify the nearest associate and provide them the location information for the customer in need, so they can respond to help.  This idea can also be used at help desks, service desks or reception areas, to monitor and alert for situations where a customer is present without an employee attending to them.


Procedure Compliance

The field of video analytics continues to evolve.  The increase in higher resolution cameras and more affordable, faster computer processors, provide a cost effective means to host more CPU intensive video analytics algorithms.  The result is the ability to recognize more complex gestures, provide better object classification and provide more big data.  In addition to more precise customer behavior, these advancements can also provide information to increase employee safety, ensure level of compliance and measure the effectiveness of training.  For example, in retail food handling, intelligent video can provide data as to whether gloves, hairnets or hats are being worn at food handling stations.  It also has the capability to provide data for the length of time that food has been sitting out of the freezer or the average time food items remain under the warming lamps.  These capabilities would not have been possible a few years back, but sensor affordability and new video processing techniques have now made it a reality.


Loss Prevention

A final video-derived data opportunity area involves combining video data with data from other systems, such as cash registers, to allow a more advanced method to evaluate and aid in loss prevention.  Notable scenarios include monitoring for a common deceptive practice whereby the clerk moves a product across the checkout scanner and covers the barcode so that the item does not get added to the total.  The video data logs the scanning transaction, but the lack of a corresponding entry in the cash register system, logs a conflict event.  A similar transaction occurs when multiple items are purchased, video is used to monitor the number of items, and in some cases the types of items.  If the number of items entered into the register does not match the number of items seen by the video analytics, a conflict is logged for further evaluation, and if desired, action can be taken in real time.



There is no “bigger” data than data derived from video.  It is a huge source of information that can be used for long term analysis, as well as, real time actions.  The accuracy of the data is of course predicated on many items including the camera resolution, its location and the quality of the video analytics solution used to process the data.  Odds are you already have the video available, you just need to take the next step to extract the big data.