It has been shown on various television legal and cop shows and is becoming commonplace in venues such as airports, institutions of the banking industry, and many more locations. If you haven’t figured out the topic of this column you should have because it’s staring you in the “face.” That should tell you I’m talking about facial recognition technology. It would seem that it is a relatively new development, but that is incorrect.
Facial recognition was actually developed in the 1960s. Woody Bledsoe, Helen Chan Wolfe, and Charles Bisson started using the computer to recognize human faces in 1964 and 1965. It seems a government agency funded the initial project and due to secrecy issues remained unnamed and that is the reason that there was no publicity concerning their breakthrough work. According to Wikipedia.com “Based on the available references, it was revealed that Bledsoe’s initial approach involved the manual marking of various landmarks on the face such as the eye centers, the mouth, etc., and these were mathematically rotated by computer to compensate for pose variation. The distances between landmarks were also automatically computed and compared between the images to determine identity.” By today’s standards that would seem relatively simple, but in it’s inception that was far from an easy task. They started with a substantial number of images and had the computer compare the landmarks between each image to determine the identity of the person pictured.
Quoting Woody Bledsoe “This recognition problem is made difficult by the great variability in head rotation and tilt, lighting intensity and angle, facial expression, aging, etc. Some other attempts at face recognition by machine have allowed for little or no variability in these quantities. Yet the method of correlation (or pattern matching) of unprocessed optical data, which is often used by some researchers, is certain to fail in cases where the variability is great. In particular, the correlation is very low between two pictures of the same person with two different head rotations.”
Bledsoe’s work was continued in 1966 by Peter Hart of the Stanford Research Institute. Christoph von der Malsburg developed a system in 1997 with graduate students at the University of Bochum in Germany and the University of California that bested the other systems at the time. The Face Recognition Grand Challenge (FRGC) in 2006 indicated that the new algorithms are ten times more accurate than the face recognition algorithms of 2002 and 100 times more accurate than the ones from 1995. Some of the algorithms beat human participants in recognizing faces and could even recognize the difference between identical twins. That gives you an idea how the technology has really improved over the short span of years.
There is another use for facial recognition technology that I found quite surprising. Would you believe it can be used in weather forecasting? A recent article in Physics.org titled “Facial recognition technique could improve hail forecasts.” The new study is from the National Center for Atmospheric Research (NCAR) and it was recently published in The Monthly Weather Review of the American Meteorological Society.
According to NCAR scientist David John Gagne who was the team leader “We know that the structure of a storm affects whether the storm can produce hail. A supercell is more likely to produce hail, than a squall line, for example. But most hail forecasting methods just look at a small slice of the storm and can’t distinguish the broader form and structure.”
Large hail can destroy crops and seriously damage structures and vehicles. The National Science Foundation (NSF), is a sponsor of NCAR. Nick Anderson, an NSF program officer is quoted as saying “Using these deep learning tools in unique ways will provide additional insight into the conditions that favor large hail, improving model predictions. This is a creative, and very useful, merger of scientific disciplines.”
This still doesn’t explain how facial recognition technology can help forecast hail. The actual answer is relatively simple. Many of as kids, and some adults still do, recline down on the ground and look up at the clouds to see what animal or object they might resemble. That is a very simplistic explanation of how the look of cloud formations can be used to predict hail production.
It takes a specific set of criteria for a storm to produce hail, particularly large hail. Quoting the phys.org article “The air needs to be humid close to the land surface, but dry higher up. The freezing level within the cloud needs to be relatively low to the ground. Strong updrafts that keep the the hail aloft long enough to grow larger are essential. Changes in wind direction and speed at different heights within the storm seem to play a role.”
David John Gagne used a machine learning computer model that was designed to analyze visual images. He trained the model by using images of simulated storms. The old way was to examine a vertical slice of the sky as with Doppler radar to determine the strength of the storm and the winds within it. This new method also looks at the “big picture” horizontally.
To better understand what a particular storm cloud produced Gagne ran the model backwards to “pinpoint the combination of storm characteristics that would need to come together to give the highest probability of severe hail.” Some of the things they learned are that storms have lower-than-average pressure near the storm top (a combination that creates strong updrafts) are more likely to produce severe hail. So too are storms with winds blowing from the southeast near the surface and from the west at the top. Storms with a more circular shape are also most likely to produce hail.”
In 2017 David John Gagne used “actual storm observations for the inputs and radar-estimated hail sizes for the inputs to train the model. He found that the model could improve hail prediction by as much as 10%.” The model has been in operation for the past several Spring seasons and he is in the process of verifying how accurate the model was during that time period. He is now working with researchers at the University of Oklahoma on the next step as they begin testing the newer machine learning model using storm observations and and radar-estimated hail. The goal is to have the model transition into operational use.
Even if this research is very successful there would be no way to prevent hail from damaging crops or property, but it would give the National Weather Service personnel a chance to issue a severe hail warning to the public so they could have at least a chance to get to a place of safety away from the dangerous falling hail.
Let me know what you would like me to talk about or explain. You can comment below or email me at: [email protected].