Machine vision
How to correctly recognize robot positioning technology (3)
Third, the "humanization" of the machine must first study the "human"
Most of the current research on "machines" is inspired by people and animals. For example, we studied the eyes of flies and then invented the camera; we studied bats and invented the radar. Therefore, in this new round of industrial revolution, the development direction of "machines" is "humanization" and "anthropomorphism", so that machines can replace people to work, and even work better than people. Of course, at present, the machine can replace humans in some aspects, and can be better than humans, but the scope of application is still relatively narrow. The immediate task is to further make the machine closer to humans.
The "anthropomorphism" of a machine does not mean that the machine looks like a human. This is a superficial understanding. The core of the anthropomorphism of the machine is that one can observe the objective world and obtain information like a human; Think the same way, make decisions; Third, the machine can be as flexible as a human to perform complex actions. Everyone who has watched the "Super Heroes" movie knows the "big white" in it. Without considering the current technology implementation, "big white" is an anthropomorphic robot.
1441787265135734.png
Some people say that "science cannot be separated from fantasy", yes, in the machine vision industry, this is more obvious. We look at Tony's "virtual stereo laboratory" in "Iron Man" and "T800 robot" in "Terminator". , "Skynet", and so on, these science fiction movies all dream of "visual technology" to a different realm. If such a level of technology can really be achieved in the future, then we think that we can study the combination of "human" and "machine" now.
When people work, study, and live, no matter how complicated the work is, they are single or combined behaviors in the following aspects:
Ø Look: Use the eyes to obtain information about the objective world and pass the information to the brain for analysis. For a robot, it is equivalent to a robot's camera. At present, in mature industry applications, most of them can only achieve two-dimensional detection (the robot has only one camera), that is, the current robot can only scan and obtain object information on the same plane. Then the future development direction is to reproduce the three-dimensional scene in all directions, which means that the robot needs to have two eyes like a human.
Ø Listening: People obtain sound information through their ears. The application of robots in this area is mainly concentrated in the entertainment industry, and it is rarely used in the industrial field. But the true anthropomorphic robot in the future, the sound sensor is a very important part.
Ø Tactile, etc .: People collect information by touching other objects or smelling their scent. The robot's technology in this area is mainly based on contact detection, and its core technology is pressure sensors. However, the information that can be obtained by this technology is single and rarely applied. When judging the condition of the object surface and the smell of color, it mainly depends on visual technology.
Ø Thinking: People obtain information through the touch of eyes, ears, and various parts of the body and pass it to the brain. The brain analyzes the shape and position of objects through years of experience, and then issues actions according to our wishes. So for robots, this includes 2 parts: First, the "signals" (the signals here may be various types, such as image signals) obtained by sensors (mainly visual sensors, but also pressure sensors, sound sensors, etc.) , Voltage pulse signal, charge signal, etc.) to process, extract useful information (that is, the basis for the computer to make judgments) and convert it into a digital signal recognized by the computer class. Then the computer makes a logical judgment on the "digital signal" according to the function of the robot, and then sends the execution command to the robot.
Ø Action: A person controls various parts of the body to perform behaviors through a neural network, such as running, frowning, laughing, etc. For robots, the action is performed by a motor-controlled actuator. The future development of robots in this regard is mainly to increase the speed of actuators, reduce inertia, and enhance stability.
1441787392549406.png
By comparing human behavior with machine behavior, we can roughly know the future development direction of robots. For the "Pepper Emotional Robot" project jointly invested by Alibaba, SoftBank and Foxconn, from a practical point of view, its significance is more strategic. There are quite a few domestic manufacturers that can develop robots on their own, but few can achieve international standards, and the level of vision technology of domestic robot manufacturers is very low. Most domestic pure vision companies rely on foreign technologies such as Cognex and Keyence. There are few companies that really take their own research and development routes, but some domestic companies are still exploring and trying. For example: Foxconn, its route is the road of "introduction-digestion-innovation and research and development"; Huawei purely relies on its own technology accumulation and continues to follow its own research and development route; Vespa image focusing on machine vision technology has also been independent research and development Despite the ups and downs, the good news is that it has finally achieved results. It started with basic vision hardware research and development, and later relied on the introduction of advanced robots to support autonomous vision software technology. The "visual robot platform" has taken precedence over the entire machine vision industry.
1441787413616316.png
In summary, robot positioning technology is actually a very broad concept, and it is not just the common two-dimensional and three-dimensional coordinate positioning. Robot positioning actually uses other sensors to provide a basis for judging the robot's actions. Among them, the visual sensor (who makes the image the carrier with the largest amount of information in the signals provided by all the sensors). The basis for studying robots or "smart robots" should first be studying "machine vision technology".
∷
湖南智能-MSN
湖南智能-Skype
湖南智能-阿里