The world at large knows Disney as the family-friendly cultural behemoth that is responsible for almost single-handedly putting animated films on the map. While people are also aware that Disney is the most successful production company in history, few realize just how big the company is, and the vast areas of work the company is actively involved in.
The Disney Research team, spread out all over the world and drawing from a global talent pool, is responsible for a number of significant innovations. The latest topic that Disney has published a research paper on promises to revolutionize the way surveillance is carried out.
The cameras you see clipped to the corners of walls in commercial spaces provide a continuous feed of their recording to the establishment’s security personnel. It’s up to the personnel to keep an eye out for any person caught in the recording who might constitute a threat to the area.
Of course, such individuals can hardly be expected to stand in one area to be continuously tracked by the same camera long enough for security to find them. The person will be constantly moving between different parts of the establishment, showing up on different camera feeds that will all need to be kept an eye on by the security personnel to track the person’s movements.
New image recognition software is being developed which will take over the tracking part of surveillance. It’s the same type of software already being used by Facebook and various other sites to confirm the identity of their users.
While current software works great when the subject is staring right at the camera and standing very still, things get more tricky when the person caught on camera is moving around. Different parts of the face get varying amounts of light at different angles. Subtle movements of the face or body can throw off the software’s recognition capacity. The Software can be confused well enough to allow the individual under surveillance to slip out of view simply due to random background objects. Or differently colored hair. Or sunglasses.
A Patchy Solution
For some time now, Disney has been working on improving the image-recognition ability of the many, many security cameras installed all over the Magic Kingdom.
The solution they’ve come up with is as elegantly simple at the theoretical level as the actual work put into it is complicated. Basically, instead of having the software identify the full face of an individual and trying to track it across different feeds, the new program identifies tiny, rectangular sub-regions of the person’s face called ‘Patches’. Identifying individual patches instead of the full face means the software has to do several times more work to process the image than it currently does.
But the software’s image-recognition ability also improves significantly. Since the software is now using a composite of different parts of the individual’s face and body, even if those parts show up under different lighting and background condition, they become far easier to identify and track.
Experiments conducted by the research team have already proven that this new system of identification works much better than earlier-used detection systems.
So how soon will we see this new and improved form of video surveillance hit the streets of The Magic Kingdom? That’s hard to say. The research that has gone into designing the software is solid, but also very recent. A host of real-world experiments will have to take place before the project is given the green signal. Which means you can rest easy for now knowing the cameras won’t be able to track you after you punch out random employees on a visit to Disneyland and try to make a break for it.
Once the software does see wide-scale deployment in the coming years, it will significantly raise the standard of satellite surveillance operations. In the past, some of the worst accidents have occurred because automated drones had been unable to properly identify and track their targets, and innocent people ended up paying the price for that mistake. This new software should go a long way towards making the work of tracking criminals easier for law enforcement, with the improved data sets obtained by the software from the video also carrying more weight in court.
Of course, there is that inevitable flip side of the research pertaining to the privacy issues that crop up whenever the topic of surveillance comes up. How soon before you get illegal drones relentlessly stalking celebs in their private lives, no matter how hard they try to dodge the drone’s camera? How soon until every single step you take outside your house is caught up on some sort of government camera that can track your movements no matter where you go or what you look like?