Today we are going to talk about it. It is a first reaction that is very surprising: "Is there such an operation?", but I think about it and I am extremely frightened.
This incident is a big news recently made by the US military. According to an article by The Register columnist Katyan Kutcher on April 17, the US Army Research Laboratory is working on a technology that will enable the identification of people in the dark through the blessing of AI's deep learning technology. The possibility of the face.
It is not difficult to get a vision at night, and the development of infrared technology is very mature. This type of technology is used for wildlife shooting, 24-hour surveillance, and more. Coupled with thermal imaging, it is no longer a problem to be able to judge in the night that you are standing in front of an individual rather than a stone.
But to achieve the accuracy of face recognition without lighting, this is a bit shocking. So, what exactly is the US military going to do?
AI+ Thermal Imaging: Not only see you, but also see youThe first thing we need to know is that thermal imaging is a technique for generating infrared image energy by non-contact detection, converting it into an electrical signal, and then generating images and temperature values ​​on the display. Thermal imaging devices are very common in military applications. For example, Apache helicopters and other ground vehicles such as armored personnel carriers are equipped with thermal imaging cameras. This ensures that people can be detected in low visibility situations, such as fog and night.
There is a problem here. Although the thermal imaging device is equivalent to giving a pair of eyes to the soldiers who are active at night, the eyes are actually double "myopia". It reveals the basic outline of the obstacles that appear opposite, allowing the soldiers to judge whether they are human or stone. However, it is limited to this. The people in front of you are enemies or friends, and there is still nothing left.
Based on this, the US Army Research Laboratory has developed a deep learning-based AI recognition system. By the processing of the convolutional neural network, the boundaries of the human eye, nose, lips, and the like in the thermal image are marked, so that the overall shape of the face can be determined. Next, the nonlinear regression model maps these features into the corresponding face appearance. This system, called "multi-region synthesis," is trained by the loss function, and the error between the thermal image and the visible image is minimized, so that a relatively accurate face image can be constructed.
But how do you know this face? After all, it is not very useful to just get an image of a human face. The main purpose of this technology is to judge who is the opposite person. This requires a database to match the instantly generated faces.
Therefore, in order to be effective, it needs to match the constructed image with known faces in the database, such as prison criminals, terrorists, etc., in order to identify the target. In other words, the facial image captured using this system must match the image in the database. If it can be applied early, the arrest of criminals in the Boston-like terrorist attack may not be so costly. In this way, the biggest advantage of this system at present is to catch fugitives.
It sounds like the future of this technology is still very broad. But before the US military is happy, some people come to pour cold water.
At the IEE Mobility Computer Vision Application Conference held in Nevada in March, some researchers believe that although the accuracy and conventional imaging accuracy of the multi-regional synthetic model of thermal imaging technology exceeds 80%, it seems that the results have certain hopes, but because of The accuracy of the thermal imaging accuracy, the face synthesis accuracy of this technology will also change. Therefore, the perfect combination of these algorithms with current thermal imaging techniques still has varying degrees of difficulty.
Interestingly, some domestic media have exaggerated their ability to translate “through the wall†when they are reporting. What we need to know is that the current thermal imager can't break through the wall to detect the radiation emitted by the human body. Therefore, the “through the wall†says that the composition of the steamed bread is bigger.
For this technology that has just made initial progress, such criticism should be considered a spur or direction. But now the problem is coming out again. Even if the technology is mature, there are not so many faces in the database to match, everything is not white?
Therefore, at present, the algorithm can temporarily solve part of the problem in the case where the existing database is combined. But if you want to achieve more accurate and wide-ranging applications, it still depends on the expansion of the face recognition database. For example, a business organization called Terrogence claims to have a database of thousands of terrorist faces. The customers of this database include the National Security Bureau, the US Naval Intelligence Agency and government agencies.
From this point of view, this technology still has great potential for pursuing fugitives, fighting terrorists, and fighting at night.
No entry: stop at the borderWe were surprised at the above, and then we are going to be frightened.
The biggest role of nuclear weapons is to be shocked, but when its enormous destructive power was used to kill people, Einstein regretted it. This also means that any technology, including the AI+ thermal imaging we introduced today, greatly reduces the threshold for face recognition, regardless of its original intention, if there is a possibility of harm to the public (especially when When the technology is mastered by the military, then someone must be supervised.
Imagine that you and your friends are walking around the street at night, enjoying a moment of tiredness after a hard day, while the infrared camera at the intersection is silently scanning and analyzing your two avatars, and at the same time, I don’t know when Collect your avatar in the database and make a judgment: security personnel.
Although it does not affect your normal life, it will give you a deep concern: our lives are constantly monitored by countless pairs of eyes. Maybe we didn't feel that we were hurt. But this is likely to mean that in this city we have lost unauthorized privacy. In this era of personal privacy that can be used as a commodity to exchange huge profits, everyone does not want to become a commodity without knowing it. It is precisely because of this that people will feel so angry about Facebook's disclosure of user data.
Therefore, regardless of the original intention of the US military's technology, we all hope that its purpose is only to deal with the dark elements of society, and the people must also hope to have strict supervision over this technology. At the same time, restrictions should be imposed on the use of government and the military, and to ensure that technology does not flow out, so as not to be used by people with ulterior motives. After all, everyone should have independent powers to their own privacy, and not be a Truman exposed to the camera.
Our wire harnesses are widely used for various parts of automobiles and truck , Like engine valve cover gasket/diesel gasket(DORMAN-FORD), headlamp(GROTE), Intelligent Anti-collision System(Mobileye), audio, rearview mirror, power seat, etc. Also the complete engine cable assembies for electric vehicle.
Related Products:ul wiring assembly,engine gasket cable,cable harness.
Automotive Wire Harness,Auto cable assemblies,OEM Automotive Wire Harness
ETOP WIREHARNESS LIMITED , https://www.wireharness-assembling.com