The use of big data has followed a noticeable pattern in its evolution, similar to that of all technologies. Its implementation has not only grown to encompass more people and organizations, it has gone from simple uses to more complex ones. When big data analytics first caught on among businesses, it was used to analyze vast sums of structured data, from financial reports to customer statistics. Analysis of structured data is relatively easy since by its very nature it is organized. Big data analytics wasn’t quite needed to perform the analysis, rather it was needed to handle the huge amount of information being supplied. As we’ve grown accustomed to analysis of structured data, we’ve moved on to unstructured data, things like images and audio. The goal has always been to get the technology to a point where video and the data it provides is analyzed at the same pace. Video in regard to big data has often been referred to as the “Next Big Thing” or the “Next Frontier.” Data scientists have eagerly anticipated the time when video was utilized to its fullest due to its promise and potential. That moment, however, has already come. It’s time to stop thinking of video as the “Next Big Thing” and understand that video’s use within the big data sphere has arrived.
First, let’s take a step back and think about the influx of video we’ve seen in just the past decade. YouTube is perhaps the best example of this. The video streaming website has over a billion users, and hundreds of hours of video are uploaded every single minute. Many people now watch YouTube regularly rather than normal television programming. And that’s just one website, though it happens to be the most popular one. Netflix is another great example of video streaming taking off at a level never before seen. Go anywhere on the internet and video can be found everywhere. That’s just a brief glimpse of the web. Cameras in general are popping up in all sorts of places. Surveillance cameras alone have reached millions in number in some cities, and that’s not counting what private businesses and individuals place on their properties. Now think about the way people can now record their own video by using their own smartphones and other mobile devices. Needless to say, it’s easier than ever to produce video content as the technology has become more affordable. More importantly, there are no signs this trend will let up anytime soon. Video will only become more prevalent around the world.
The true excitement that comes from video being used for big data analytics is how valuable it can be as a data source. While its unstructured nature represents its complexity, that also means video contains a great deal of information that can be analyzed and applied. Sales data, just as an example, shows fairly basic information (who bought an item, which item was bought, etc.), but video data reveals so much more. The potential is downright staggering at times. Thanks to advances in intelligent video analytics technology, video can be analyzed all in real time, helping you know and understand what is happening right now. There’s no need to wait hours, days, or weeks for the information you’re seeking to be pumped out of an algorithm. You get what you want in the moment, allowing you to respond in real time.
Of course, this brings up one of the biggest challenges related to using video for big data purposes: extracting that data. It’s much easier to say that video can provide insightful details, it’s another thing entirely to actually perform that function. For obvious reasons, extracting and recording that data manually is simply not feasible. Even a modest retail store will use a dozen or more cameras, and expecting a human to monitor all of them at once, 24 hours a day, and getting any meaningful data from them is expecting far too much. Systems and software are needed to identify what’s being recorded in the video and determine what information can be gleaned from it. While this is certainly a challenge, it’s not insurmountable. In fact, many organizations have already started to do just that.
Video analytics is not some conceptual idea that data scientists hope to use someday in the future; it’s here and it’s already providing valuable insights. But in order to use video analytics, developers needed to create computer vision algorithms, effectively developing a system that can analyze everything about a video instantly. That all comes down to analysis of individual pixels, where they fit in the bigger picture, and even the context of the video. Video analytics has come a long way in a short amount of time. Applying new technologies, it’s now possible to identify objects within a video, track their movements, and even determine specific types of behavior. This might be something like a vehicle in a parking garage, or it could be a customer checking out displays in a retail store. Video analytics has even advanced to the point that hundreds of objects within the frame can be analyzed, essentially mined for data. If it’s recorded, chances are data is being generated.
All of this has been made possible thanks to technologies like machine learning and the beginning stages of artificial intelligence. It would be almost impossible to program a system to categorize every type of behavior and identify every type of object. The beauty of machine learning algorithms for use in video is that the algorithms are always learning and improving. Based off of user feedback, the system will constantly update, getting better at mining data from video. That’s why we’ve seen such progress being made in things like speech and facial recognition. These systems attempt to act similar to human brain functions, allowing them to find faster connections and become better at what we want them to do.
It’s important to recognize that while incorporating video with big data solutions sounds like a complicated endeavor reserved only for the largest and most tech-savvy companies, businesses of nearly any type or size can now use it. In fact, one industry that has grown very fond of video analytics is the retail industry. No strangers to setting up surveillance cameras, many retails companies have combined their existing cameras with new big data analytics capabilities. By using video analytics to observe customers in their stores, they can find out much more information about who is shopping there. It’s similar to the way Amazon collects information on customers as they browse through their online store, only this potentially gives even more information. Based off of data analyzed from their videos, retail outlets can determine traits like age, race, and sex of each customer who passes through their doors. They can track their movements and analyze behaviors, even to the point of determining which items their eyes are drawn to and which areas of the store they spend the most time in.
Based off of this information, retail stores can gain more detailed information about their customer base. They can even determine how best to manage high traffic areas and where to place displays for maximum exposure. From these videos, they may even be able to keep track of their stock and know when to replenish items before inventory is taken. Part of the reason retailers and similar business have access to this technology is that it is available through cloud services. That also means employees don’t have to be big data experts to use such systems.
These ideas can be applied in many other areas. Smart cities are looking to use cameras to better manage their cities. With these cameras, they can collect data on traffic patterns and find new ways to manage stoplights and streets in order to allow for smoother traffic flow. Surveillance analytics may also be employed in law enforcement and waste management. Due to the prevalence of cameras and use of predictive analytics, problems may be detected before they even occur. London, for example, has made a concerted effort to install cameras all over the city, with some estimates reaching 5 million. Only through computer vision algorithms can the data being recorded through these cameras be analyzed and mined for more insights.
Video has become widespread in nearly everything we do. While it seems like a tall task to analyze everything being recorded, that capability is already at our fingertips. There are many applications for video in big data, from better understanding of customers to ensuring effective city management. Even more exciting are the possibilities that come from a more connected world, particularly through the Internet of Things. Advanced technologies and concepts have made this all available, leading to an exciting future. While many still seem to think of video used for big data is something still to come, its uses are already being put into practice. There’s no need to wait for what’s on the horizon, because the applications are here. Video truly isn’t the next great development in big data analytics but rather an evolution that has already occurred. Businesses and organizations should look to take advantage of it right away.