戍天九思原語
Recently, artificial intelligence has made a major breakthrough in machine vision, giving the machine a "sixth sense" superpower, application The prospects are amazing and deserve great attention.
The shadow camera has predictability
On October 27, 2019, the Massachusetts Institute of Technology published a new study, The study Let the machine predict the trajectory of the object behind the obstacle by analyzing the changes in light and shadow in the surrounding environment.
This is based on a system called a shadow camera, which can determine whether an object is approaching by analyzing the changes in the intensity of light and shadow in the continuously captured images.
In this study, the experimenters used the "shadow camera system" on turning cars and autonomous wheelchairs in hospitals. Using the traditional lidar system , the reaction time is 0.5-1 second faster.
Don’t underestimate the time of 1 second, is equivalent to making the autonomous car as predictable as an experienced driver, can be 1 second ahead of time Discover the signs of prevention, which can effectively avoid accidents caused by obstruction of sight. For example, a self-driving car can be used to determine whether there is a pedestrian crossing the road behind the bus next to it, so as to take evasive measures in advance.
Artificial intelligence practiced "perspective eyes"
On September 27, 2019, a study by the MIT research team greatly improved the machine’s "perspective" "Eye" ability. In this study, the machine can determine what actions the people behind the wall are doing through a wall. For example, outside a room, the machine can see that two people in the room are shaking hands. It can also be seen what kind of posture the caller is in a closed phone booth.
▲Comparison of real pictures and results in the MIT team experiment
What is the principle? As we all know, although visible light cannot pass through walls, light is also an electromagnetic wave, and most electromagnetic waves are not blocked by walls. Wi-Fi at home is a kind of electromagnetic wave, no matter which room you are in, it can pass through the wall to reach your device.
In this experiment, the machine can obtain a perspective eye because it actually monitors the electromagnetic waves transmitted from the object. In the past, scientists have done similar experiments, but due to the low resolution of imaging technology, large errors, and complicated reflection environment, the actual imaging effect has been poor, and the captured images are invisible to the human eye.
However, the MIT team used a new idea for research. They compared the electromagnetic wave image taken through the wall with the normal image taken by the traditional camera through , and then Use machine learning to find the corresponding law. In this way, the machine gains the ability to understand the signal through the wall.
Artificial intelligence makes white walls a "mirror"
In the future, maybe you take a photo against the ceiling of the classroom and you may be able to see the whole thing. The answer on the class paper. Because even a large white wall contains very rich information. This large white wall reflects the light and shadow of many different objects.
Through machine learning technology, these light and shadowAfter analysis, it is possible to restore the objects in front of the wall. Therefore, scientists can turn a white wall into a "mirror". This is also a research result published in the journal Nature in 2019.
The "supervision" technology has a broad application prospect
It can be seen whether it is the foresight ability of the shadow camera, the "perspective eye" of the machine, or the wall change The "juggling" of the mirror is essentially the "superpower" that allows artificial intelligence to analyze the subtle and complex hidden information in the environment, thereby developing the "superpower". These "super powers" of seeing the micro-knowledge will reshape our information environment, privacy environment and living environment in the future , and will surely bring about revolutionary changes in the battlefield environment.
At present, self-driving cars have a fatal weakness: due to technical bottlenecks such as the delay of communication signals and the delay of automatic control systems, they cannot handle sudden emergencies like pedestrians crossing the road. Emergency, this is the root cause of many fatal accidents in Tesla's self-driving cars. With this "super vision" technology, not only can you see more clearly, but also have a 0.5-1 second foresight ability; and with the millisecond delay 5G technology, you can basically realize real-time automatic Control, such problems will be solved.
At present, military robots also have technical weaknesses similar to those of self-driving cars. The most advanced robot of the Russian army is not autonomous combat, but remote control combat. In December 2015, in the battle for the high ground in the Syrian battlefield, the Russian army put in a robot battle group to cover the Syrian ground forces to attack the position held by the "Islamic State" militants. The militants collapsed on the front in only 20 minutes, with 70 Many people were wiped out, but only 4 of the Syrian army were injured.
▲"Platform-M" crawler combat robot
If there is this kind of "super vision" technology With 5G technology, combat robots can truly realize autonomous driving, have foresight and perspective capabilities, and robot troops will be even more powerful and truly achieve: first enemy discovery, first enemy lock, first enemy fire. No wonder Russia recently announced that it will form a combat robot force before 2025.
▲"Uranus-9" battle group
It is not difficult to foresee, as this "super-vision" technology matures, It is also possible to develop a new generation of highly sensitive military reconnaissance and situational awareness equipment, making more impossible possible .
Remarks: The picture comes from the Internet
