A Revolution in Camera Resolution
Most of us have seen the footage from convenience store security cameras … usually in the context of a police report or plea for citizen help. The videos are grainy, have poor color, and often lack the detail necessary to see the suspect clearly, let alone identify them in court. However, those systems bear no relation to the new generation of surveillance cameras that have entered the market in the past 5 years. The following progression (from the Avigilon website) shows the amazing improvement that has occurred (the red outlines and dates are mine).
Every serious security system vendor now offers cameras at the 4K level. Avigilon seems to be leading (for now?) the race to introduce cameras with the highest resolution. Their 7K surveillance camera is commercially available and captures an image more than 7,000 pixels wide!
More technology is always cool … but is it really useful? The security industry seems to think so, but I sense a bit of uncertainty about whether the really high resolution cameras will become more than a niche play. For one thing, a 7K camera emits a massive data stream. If you want to store that stream for any length of time, your networking load and storage investment will go up very quickly. History tells us that eventually the bandwidth and storage will be catch up and the camera costs will come down, so I am sure these will become common in a few years. But for now there are other ways to achieve most surveillance goals.
However, I am beginning to suspect that the value proposition for these higher resolutions may be stronger for systems that watch and analyze production operations. There, the additional detail should enable us to create distinctly new management capabilities … with direct impact on production flow, yields, losses and other critical bottom line issues. The remainder of this article explains why I think this might be true.
Higher Resolution Extends our Vision
There are many aspects of the typical factory, warehouse or industrial facility that are hard for a single human to view with their naked eye. The following diagram illustrates the range of physical scales that we typically encounter. It goes from objects too small for the eye to resolve up to panoramas and vistas that are too big for the human eye to fully take in. As humans, we can see (and focus on) things in the middle of the range. These are the human-scaled activities that are eyes are designed to view.
These limitations directly limit our ability to manage complex processes and operations. If you think about it, many of the standard tools in the lean manufacturing arsenal were at least partially invented because of this limitation. Why do we need Andon (visual) signals on our lines? Is it because we can’t see everything at the same time and we need something to draw our attention? Why do we do bottleneck analysis? Is it because we don’t have the ability or patience to watch the various backups as they grow? For what its worth, I think the new generation of video technology might offer some other options. We shouldn’t abandon the proven methods, but surely they can be enhanced if we can see more clearly and revisit our past more thoroughly and accurately.
High Resolution for Wider, Clearer Vision
It’s human nature to think that we see everything in front of us. However, the human eye has a characteristic called foveal vision. Our eye can only focus on a small part of the panorama in front of us. All the rest is peripheral vision. The diagram below on the left is a rough approximation of our capabilities … they vary somewhat between individuals. The photo on the right was taken from my desk and the red circle shows the part that that seems fully distinct when I look at the lemon tree.
Consider what this means to workers, supervisors, technicians, engineers and managers when they watch activity on a factory floor or in a distribution center. Even if they are watching continuously, they can really only register a tiny portion of what is in front of them. Now consider a video camera. It records every pixel over the entire image with equal fidelity and it can retain that fidelity as long as you record the video stream. You can return at any time and view the image and every part of the scene at identical levels of detail.
Now consider what is becoming possible with the new, high resolution cameras. The following examples come from the Avigilon (www.avigilon.com) web site. You have to open the linked page and look for the demonstration image (like the one shown on the left) a bit down the page. Since no web browser or monitor can do justice to a 4K … let alone a 7K … image, you must hover the cursor over the image to see a small window with the full camera resolution. Move the cursor and see how much more detail you can discern. A high resolution camera can record that level of detail across its entire image and every image can be saved for viewing at some arbitrary later date.
|Avigilon 4K example
Avigilon 7K example
A High Resolution Overview of Operations
My concept is to use this capability to make a complete recording of all facility operations. To do that, mount several high definition cameras around the periphery of the work area … two at a minimum. As the following diagram shows, the cameras should view the space from opposing vantage points. The key is to overlap the fields of view. A given activity might be hidden from one camera, but the chances of it being hidden from all of the cameras is much less. Add another camera or two to improve the coverage and see better into potential blind spots.
A simple system like this promises several intriguing possibilities:
- The system can record a complete history of activity in the facility … perhaps not every detail (an object might block the view), but certainly all major actions and occurrences. It may not tell you why or how something happened, but it will probably establish what happened, when it happened and what was going on nearby at time of occurrence. In other words, it can give a powerful visual context to everything that has recently happened in the facility.
- The system should be able to see and record the movement (flow) of all people, equipment, and product in the facility. Understanding process flow is often the most difficult thing that process managers have to do. I am especially intrigued with the possibility of using analysis tools like Dartfish to mark, track and analyze key movements and flows in the facility. How might this help when you are trying to improve spread-out activities like shift changes?
- The system should generate a wealth of raw material for training and for remote problem solving. If there are “interesting” events that you would to prepare your staff to handle, chances are they will have appeared somewhere in the video record. Find them, export the video (possibly from several angles), and show it to your employees during training. Give them videos of real situations to discuss and understand.
- The extensive, high quality video record of operations can be made more useful and searchable by adding meta-tags … either automatically from the security technology or after-the-fact with post-processing in analysis tools like Dartfish.
Thanks to ruthless competition in the security industry, systems like this are falling rapidly in price … especially compared to the potential gains in flow efficiency, reduction in loss and improved staff skills. You can buy a high quality 4K IP camera (indoor use) for $900 to $1300 (see this Bosch example). This is where I suspect that high resolution (4K at least, and perhaps 7K) video may find a large and lucrative new application.
Of course, if you want to apply video in continuous improvement, the high level overview is important, but far from enough. You will want to deploy other cameras of various types to watch more localized points of interest. Nonetheless, the output from those cameras will be greatly enhanced if it can be viewed in the full overall context that this type of system would provide.
Can you help?
I’m not an employee, investor or reseller for Dartfish, but they have kindly let me play with their software to explore any non-sports uses I can cook up. I am similarly not an employee, investor or reseller for Avigilon or any other vendor. I am willing to play with videos sent to me by others as long as they don’t hold me to a deliverable timeline and they give me permission to post useful pieces on the blog.