Unscramble the category and application principle of robot vision

In our daily life , As the equipment gradually becomes an integral part of us , We have seen that if we don't have enough visual ability , More and more applications will fail , These include aerial drone collisions and robotic vacuum cleaners “ eat ” What they should not have eaten . There is a lot of room for the development of service robots , But to develop service robots , It is necessary to let robots acquire human capabilities , The first is vision . Vision is very important to people , Human beings get information 90% The above depends on the eyes , Let's take a look at the frontier of artificial intelligence —— Machine vision . Technological evolution and innovation , Promote the reform and progress of the manufacturing industry
Intelligent 、 Bionics is the highest stage of industrial robots , With materials 、 Control and other technologies continue to develop , More and more laboratory products are commercialized , Gradually applied to various occasions .

With the development of Internet of things , Multisensor 、 There will be more and more precision industrial robots with distributed control , Gradually penetrate all aspects of manufacturing , And transform from manufacturing implementation type to service type . Tactile 、 Industrial robots with force or vision , Able to work in complex environment ; For example, it has recognition function or further increases self adaptation 、 Self learning function , That is, to become an intelligent industrial robot .

Robot vision , As AI( Artificial intelligence ) A fast-growing Branch , The goal is to give robots the same vision as ourselves , In the last few years , Because researchers use specialized neural networks , To help robots recognize and understand images from the real world , Robot vision has made great progress .2012 Year is a starting point , Although computers can do everything now , Recognize cats from the Internet , To be able to be in a group of photos , Recognize specific faces , But there is still a long way to go . today , We see that machine vision can leave the data center , And apply to everything from autonomous drones to robots , Can tidy up our food .

In order to better understand robot vision , A common analogy , Robot vision and human vision , Just like birds and planes flying in the sky . Both will ultimately depend on basic physics ( Such as Bernoulli principle ), To help them fly into the high air , however , This does not mean that the plane will flap its wings to fly . Just because people and machines may see the same thing , And the way to interpret these images , There may even be some commonalities , The final result may still be very different .

Although basic image classification has become easier , however , When it comes to extracting meaning and information from abstract scenes , Robots are facing a series of new problems . Illusion is a good example , Robot vision still has a long way to go .
Everyone may be familiar with the two silhouettes facing each other , The classic illusion . When a person looks at this image , They are not limited to seeing abstract shapes . They insert more backgrounds into their brains , Enable them to recognize multiple parts of the image , See two faces or a vase , In fact, all of them come from the same image .

When we pass a classification , You can manage these same images ( You can find some free ones on the Internet ), We soon realized , For a machine , To understand these complex things , How difficult it is . A basic classification , I didn't see two faces or a vase , But to see something else , Like an axe 、 hook 、 Bulletproof vest , Even a wooden guitar . Although the system is recognized as uncertain , In these images , In fact, anything can happen , It shows how challenging , Human beings still don't understand , Not to mention robots .

If we see something more complicated , This problem will even become more difficult , For example, a painting by Fukang Doolittle , Although everyone who sees this picture , May not be able to find , In fact, everyone's face is on this canvas , They saw almost immediately , More pictures than they can see .

To understand why this is such a big challenge , You need to think about , Why is vision so complex . Just like these images , The world is actually a very chaotic place . Browse the world , It's not like building an algorithm , Then it is so simple to analyze the data , It requires us to base on the actual situation , We can take corresponding action experience , And need to have a deep understanding .

in summary , Robots and drones face countless obstacles , It may be out of routine , And find out how to overcome these challenges , Those who hope to realize the AI revolution , A big problem to be solved .　 With the continuous adoption of these technologies , Such as neural networks and dedicated machine vision hardware , We are rapidly narrowing the gap between human and machine vision . One day in the future , We are even beginning to see the visual ability of robots , May surpass ourselves , Enable them to complete many complex tasks , And our society will operate completely autonomously .