How Cold, Lifeless Machines Can “See” Our Colorful and Complex World?

Nearly 80% of the information we receive about the world around us comes from vision. As humans, we have long relied on our eyes to perceive and understand the complex world in which we live. Over time, we have become adept at gaining inspiration from our observations and applying it to our inventions and creations. For decades, countless researchers have attempted to give cold, lifeless machines "eyes" that can perceive their surroundings. And so the concept of machine vision was born.

Opening the door to a new world of vision, filled with endless possibilities

At Dartmouth in 1956, several scientists met to discuss a topic that seemed almost unbelievable, even unimaginable, at the time. That topic centered around the question of whether machines were capable of imitating human learning and other intelligent behavior. As we all know, many of mankind’s greatest inventions began as pie-in-the-sky fantasies. This one was no exception. This brilliant meeting of the minds paved the way for the future development of machine vision.

In 1969, several scientists at Bell Labs created a device that converts photons into electrical pulses, known as the charge-coupled device (CCD). The device lit the way for high-quality digital image acquisition technology. Its ability to convert images directly into digital signals for storing in a computer, where those images can be further processed and analyzed, made it possible to give machines “eye implants.” It is fair to say that the invention of CCD was the launchpad of machine vision – the embryonic stage of machine vision technology.

In quick succession, DALSA, Kodak, E2V and other next-generation imaging enterprises sprung up like mushrooms after the rain, driving the growth of the machine vision industry. The year 2016 saw a series of groundbreaking events: AlphaGo defeated the top human Go player Lee Sedol, and deep residual learning and residual nets (ResNets) became the standard algorithms in the field of vision, marking a new era of artificial intelligence. The floodgate was opened. AI-enabled machine vision became increasingly popular in smart manufacturing applications, and new opportunities abounded.

Helping machines open their “third eye”

Having passed from black and white to color, from low-res to high-def, from static images to motion pictures, machine vision is now entering the visual revolution "from 2D to 3D." In the recent past, given the limitations in image capturing and data processing technology, machine vision systems had always developed around 2D vision technology, naturally making it difficult to fully reproduce 3D scenes. In comparison, 3D visual perception technology collects the 3D coordinate information of each point element in space, which allows the terminal to obtain massive amounts of accurate 3D information and obtain 3D imaging through algorithmic recovery intelligence. Not easily affected by the complex external environment and lighting, this technology is able to “open the eyes” of machines, helping all kinds of computational devices to better understand the three-dimensional world.

Currently, the ToF (time of flight) method has become one of the recognized high-quality solutions in the industry. ToF sensors come with a miniature laser that emits infrared light. The system calculates distance based on the time difference or phase difference between the emitted and reflected light beams, and combines the formed depth information with conventional camera photography to present a three-dimensional topographic map with different colors representing distance. The natural properties of the laser dictate that ToF sensors are long-range, accurate, fast, and cost-effective. Consequently, ToF sensors are capable of detecting near and far objects of all shapes and sizes, independent of humidity, air pressure, and temperature, and are suitable for both indoor and outdoor use.

With the rapid development of the underlying components, core algorithms, and other technologies, 3D vision perception technology is gradually expanding into the consumer sector. Google announced its ProjectTango in 2014, a tablet and SDK based on iToF technology that provides users with motion tracking, depth perception, area modeling, and other features. Smartphones from Huawei and Meizu equipped with 3D vision sensors have also entered the market. And, DJI's new generation of products, equipped with binocular vision systems that can sense obstacles through image ranging, are now within the reach of regular people. In the era of artificial intelligence, machine vision is continuously evolving and improving. The mechanical “eye” is heralding a revolution in perception where it may well become humankind’s “third eye.”

According to the global report on machine vision published by Yole Développement, a French market research and strategy consulting firm, the global 3D vision perception market will reach US$15 billion by 2025, for a compound growth rate of about 20% from 2019 to 2025. After decades of fluctuations, machine vision technology has now begun to flourish. To meet the needs of a wide range of applications, 3D vision perception technology is becoming increasingly diverse and comprehensive, giving rise to smaller sizes, higher performance, lower cost, and lower power consumption.

With these developments and more in the pipeline, it looks like our “third eye” is opening. And the future looks brilliant.

Coda

The industry generally believes that the development of “smart” technology is a revolution, as is about to be seen in AI vision. In the smart era, where everything is perceived and everything is interconnected, machine vision has become a fundamental ability for the evolution of the industry, and is absolutely essential for its future. Similar reports show that machine vision accounts for more than 40% of all kinds of smart applications, and 60% or more in vertical fields, such as security. In the wave of AI, IoT, cloud computing, and smart sensor technology, machine vision may become humanity’s “third eye”, giving people a glimpse into the yet unseen future world of “everything is connected, everything is smart.”