Most of technological developments have been improved first of all for military purposes and often used as/for weapons. The reason is in most of the cases is the high cost of researches, experimentations and first prototypes, which usually can be covered only by governments funding for military purposes. Thus, weapons are incredibly developed, but current issue in big discussion regards their automation. Nowadays most of the process, in almost every field, are being automised, but in the case of weapons, the question is much more delicate. Their scope is to kill and it’s important that such drastic task stays under a meaningful human control. Indeed, an International Humanitarian Law claim as illegal completely autonomous weapon.
Facial recognition and surveillance in general are concepts quite related to weapons. If seen strictly from a military prospective, the latter is an attack system, while surveillance is a defence system. This is obviously a rough simplification, surveillance systems, such as facial recognition can indeed be used also for identification of a target and, in fact, as a part of weapons.
However, the point on which I’m interested regards the automation of the two process, which is equally thorny.
FACIAL RECOGNITION AND FACIAL DETECTION
To begin, it’s important to understand the difference between, facial recognition and facial detection. While the first one is a process able to recognise a face and therefore it’s used as surveillance system. The latter is a much simpler technology, present on most of the cameras, which can detect the shapes of faces, recognising a certain object as a face. The two technologies are often combined, facial detection is indeed, the first step in automate facial recognition.
While facial detection is a completely mechanical process, facial recognition to work must be connected to a database, it has to refer to something else, would be impossible to REcognise without knowing in first place.
A facial recognition system may be programmed only to recognise ethnicity, age range and gender of a person. In that case would the process would be semi-automated, since it will be divided in two phases: a first one fully automated and a second one completely manual. The system would be much slower, but more reliable, as it would be like doing an advance search on a browser, humans select the search criteria, skimming the research, but still having a broad decisional power. More details we add to our research, more we refine it and obviously, more automated and quick the process would be. To reach a perfect automation of facial recognition systems, a very rich database must be provided, more information must be accessed (the so-called Big Data), which means people privacy must be extremely endangered or even completely violated.
An ideal facial recognition system would have in its database all the faces of living people, on regularly basis updated to verify their changes (i.e. the growth of children).
Most of the people would find abusive and potentially harmful that all their data must be stored in a database without their concession (even though, this is certainly already happening). Overall, surveillance is always a tricky issue, people like security and crime-free living, but at the same time if they must sacrifice their privacy for the sake of it, they are seldom willing to do it.
THE SURVEILLANCE PROCESS CANNOT BE COMPLETELY AUTONOMOUS
Besides the privacy issue, there are other aspects which make completely autonomous surveillance systems dangerous and imperfect.
First of all, because as I mentioned before, a surveillance system cannot certainly be considered as a weapon in the proper sense of the term, it doesn’t shot or harm directly the subject, but its effects have decisive impact on the target. Sometimes it is part of a weapon, an operation or a system which would then harm or even kill the identified target. Currently, humans trigger the process, programming machines for a certain target, and they are at the end of it, only a human verification will actually confirm that the target corresponds to the subject of the research. Whereas, if this process would be automated as well, it would be uncontrollable and potentially dangerous.
Secondly, at this stage surveillance systems are far away from being perfect and therefore would be completely wrong to let them take initiative without a meaningful human control. However, even considering their further development and assuming their perfection, they will always be possibly deceived by a camouflage. Still in the case of facial recognition systems, it’s possible to 3D print perfectly faithful facial masks; movies and TV show us everyday which miracles can be done with latest make up techniques and so on.
From all appearances, deception has always been critical to daily survival—for human and non-human creatures alike—and, judging by its current ubiquity, there is no end in immediate sight
Roy Behrens, Camoupedia
For this reason, surveillance system requires always an human intervention, even if only partial or in the final phase. Humans must control that all information are coherent, it has to apply its deliberative reasoning. The founder of FaceFirst, Joseph Rosenkrantz (already cited in my previous post Facial Recognition) opens up to a future where a combination of many biometrics systems could improve identification systems to an almost perfection level, but even there the last word must be up to a human.
“Imagine a world where you brought in facial recognition, iris recognition, gait recognition, the way you walk and fuse all those biometrics together, so image is subjected to multiple different types of video analytics, and each analytic creates a vote of confidence on whether the person is, who in fact, is supposed to be”
Joseph Rosenkrantz, CEO and founder of Airborne Biometrics Group, Inc.
With all this I don’t want to say that I’m against autonomous surveillance systems, I only claim that an human intervention is needed at some point. Autonomous surveillance systems are surely great tools to safeguard our security. Besides the already cited FaceFirst, another excellent technology is AWARE (Automated Warning and Response Engine) from Abeo, a gigapixel camera which replaces the traditional banks of screens with a single view, combining the images from many cameras to give a picture across the entire area under surveillance. It makes it possible to zoom in to a single room, or zoom out to a bird’s-eye view. And instead of relying on humans to scan for potential threats, the software will actually analyze the video itself.
– The Future of Surveillance? When Automated Brains Keep Watch 24/7 -http://www.popularmechanics.com/technology/security/how-to/a5806/future-of-surveillance-cameras/