Video Analytics and Big Data: How to Make Sense of Massive Amounts of Information

big-data-analytics-1.jpg

It’s difficult to articulate the amount of data gathered every day.

“Our accumulated digital universe of data will grow from 4.4 zettabyets to around 44 zettabytes (44 trillion gigabytes.) [by 2020]”1

Statistics like these can be hard to comprehend because we (humans) have a hard time understanding and handling excessively large numbers. But these are the massive numbers used to express the impact of big data.This data accumulation is not an abstract exercise — it has significant implications on our everyday lives. Billions of connected devices and sensors collect information on everything from our diets, medical conditions, physical activity, and driving habits, to our musical tastes and internet behaviour. This data has the potential to improve our world. But right now it’s piling up at an alarming rate, sitting useless until converted into actionable insight.

So how do we get insight from data?

boston-1.jpg

Consider the Mayor’s office in the city of Boston; they created an app called Street Bump, which automatically records road hazards (like potholes) using smartphone sensors. The city can prioritize and maintain its roadways through the crowdsourcing of insights from thousands of people in real time.

Rolls Royce uses big data in a different way. They produce enormous engines for airlines and the military, equipped with sensors that relay every detail of operation, in real time, to engineers to decide how to manage potential issues.

All this information can help humans make more informed decisions. And its use in the security industry is no different.

How Does Big Data Affect the Security Industry?

Right now, camera manufacturers are in a race to the bottom; higher image quality at decreasing prices. Having more cost-effective cameras is also making it easier for companies to add more video cameras to their surveillance systems and improve their overall security.

And why not?

More cameras means more coverage by reducing blind spots. You get a better line of site on your property and assets. However, the use of more cameras means more video data needs to be managed.

Consider an airport with a deployment of 1,000 cameras. If all cameras are recording nonstop, it will generate 24,000 hrs of video in a day. With modern compression schemes like H.264 and network storage devices, you could record the full 24,000 hours of video at a resolution of 720p and 5 frames per second with a mere 5 TB of storage/day or 150 TB if you need to store the footage for 30 days.

But what do you do with all that footage? What should you keep and what should you delete? How can you leverage this data for your business and security needs?

This is where video analytics come into play.

How Video Analytics Helps Manage Data

Video analytics can extract useful information from surveillance camera footage. The most basic form of video analytics is motion detection. Many surveillance cameras monitor restricted areas where, for the most part, nothing happens. But sometimes there are people, vehicles, or other objects in motion.

If you can determine when there is motion, you can determine when the video is important and when it’s not. Motion detection functionality now comes with most cameras.

What else can you do with analytics?

Depending on your goals and objectives, there are a number of things you can do with video analytics. For example, you can count the number of people or objects in a scene. Counts give information on how many people came or left through a door. If you look at counts over a week or month, you can recognize busy times for a retail location.

A crowd detection analytic will give you occupancy and capacity percentages to help manage large groups of people. This is different than a people counter which literally counts the number of people who cross a line to give you specific numbers.

You can also match the face of a person from store entrance to store exit using facial recognition software. Once you have a match, you can identify how long they’ve spent in the store. Given enough data you can track enough individuals to get average time people spent in the store. The average time spent in the store is a useful metric in evaluating new marketing strategies, new product offerings, and assessing various markets.

Video analytics allow you to do layers of analysis of your video data to identify patterns or anomalies that can improve security, enhance operation, or simply help you make stronger business decisions.

This analysis is similar to how data analytics are used by many other industries:

  • Credit card companies can identify unusual buying patterns to detect potential credit card fraud
  • Email providers can automatically sort email as junk, promotions, or updates
  • Search engines (like Google) can determine what internet pages are most relevant
  • Online shopping services can accurately suggest products of interest
  • Media streaming services can highlight movies that align with your tastes

But there’s a key difference with analytics in the security industry:  Processing video is more difficult than processing text or numbers. As a result, video analytic accuracy is still not as high as we would like it to be.

Big Data and Video Analytics Accuracy

Algorithms to process and understand the contents of videos are very complex and are still very much in their infancy. As a result, the accuracy of many of these video analytics is not very high. However, even with these analytics, you can leverage information from the terabytes of video you record in a day.

Consider the example of getting the average time of individuals in a store, discussed in the last section. You can’t get timing of every individual entering your store with a high confidence facial recognition match. However, even if you can identify 80% of the people with high confidence, that provides enough information over weeks and months to get a reliable average time of how long individuals spent in the store.

Furthermore, as more cameras capture more video, we’ll see analytics for summarizing and getting information from that data improve. The information being collected is also supporting a more integrated approach to how security systems are working with other physical security technologies.

Integrated Video Surveillance Systems

Security deployments are more active than ever before. Integrations give end users more information about their environment. You can interact with, and control, different deployment components or areas of the business in general, such as Point of Sale (POS) systems in a retail environment. 

POS systems can integrate with security cameras to automatically match transactions with surveillance footage.  Why is this important? Large amounts of POS transaction information are already used to study buying patterns, like which items are bought together. This information helps for marketing, and store layout planning.

By integrating POS with your video security system, you can get additional information on individuals. For example, facial recognition can tell you if two transactions were made by the same individual. This can lead to information like how often people come to the store and what they buy. You can also integrate access control systems and video management software for greater control over restricted areas.

Consider integrating a facial recognition system to an already existing access control system. Over time, given enough data, the facial recognition system can automatically determine the face of the person using a particular access card. Once it has learned the face of that person, it can automatically alert security if someone else uses the same access card.

Integrations can be powerful and give more control to operators and end users. And when combined with video analytics, these systems — security and business alike — become even more dynamic.

We’ve already seen improvements with video analytics; automatic license plate recognition technology, when used properly, can now achieve highly accurate results, and facial recognition technology has become more accurate in the last five years. So, as technology advances, bigger developments are coming.

We’ll see leaps in accuracy and the type of information that video analytics can gather; good news for all the surveillance footage your cameras are gathering every day.

Aimetis Video Analytics White Paper

1 https://www.forbes.com/sites/bernardmarr/2015/09/30/big-data-20-mind-boggling-facts-everyone-must-read/