Tech Insights: Machine vision hits new heights with AI

industrial processes calls for raising the bar in order to meet analysis challenges. Already widely used in vision-guided robotics, metrology, identification and inspection, newer vistas are being pursued, such as predictive maintenance, calling for deeper learning algorithms. Combined with machine vision, AI can be used to tirelessly detect microscopic imperfections in products and inspect minute components. Industrial robots have been around for a long time, but they are becoming more competent. Now they are entering new roles of being responsive to their environment, making judgments, and working in tandem with humans. Here is some recent related news in the field:

RFID robot tracking enters the game

Researchers at MIT have developed a system that uses RFID tags to help robots pinpoint moving objects with unprecedented speed and accuracy. A paper presented at the USENIX Symposium on Networked Systems Design and Implementation claims that the robots can locate tagged objects within 7.5 milliseconds, and with an error of less than a centimeter. The system, TurboTrack, attaches an RFID tag to any object. A reader then broadcasts a wireless signal that reflects off the tag and other objects, then returns to the reader. An algorithm sorts through the signals to elicit the tag’s response. Final computations are enacted to increase accuracy. The researchers claim that TurboTrack could replace computer vision for some applications, as it can ID targets in cluttered environments, and even through walls. “If you use RF signals for tasks typically done using computer vision, not only do you enable robots to do human things, but you can also enable them to do superhuman things,” said Fadel Adib, principal investigator in the MIT Media Lab, and founding director of the Signal Kinetics Research Group. “And you can do it in a scalable way, because these RFID tags are only 3 cents each.”¹

SETI searches for alien signals with machine vision

In the search for intelligent life on other worlds, SETI has applied machine vision and deep learning methods to create signal classifiers, in an effort to more efficiently sort through radio signals reaching this planet. The organization, which monitors electromagnetic radiation for signs of transmissions from civilizations on other planets, uses “two-dimensional spectrograms of measured and simulated radio signals bearing the imprint of a technological origin,” according to a SETI report. SETI studies use archived narrow-band signal data captured from real-time observations with the Allen Telescope Array and a set of digitally simulated signals designed to mimic real observed signals. The resulting 2D spectrogram is treated as an image, allowing parametric and non-parametric analysis to achieve high levels of discrimination and accuracy, and reduce false positives. “The most successful algorithm used a two-step process where the image was fist filtered with a rotation, scale and shif-invariant affine transform followed by a simple correlation with a previously defined set of labeled prototype examples.”²

Machine vision invades farm life

The United Nations recently predicted that the world will require 60 percent more food by the year 2050. In trying to prepare for that demand, the Centre for Machine Vision at the Bristol Robotics Laboratory, in Bristol, UK, has been working on more-efficient weeding in pastures. Melvyn Smith, Centre director, said the goal is to stop the overuse of herbicides. A camera on the back of a trailer identifies the weeds and uses a device similar to an inkjet printer to pinpoint sprayed weed killer on the weed itself. Smith explained the use of deep learning in the project: “We trained the system on what the weeds looked like in different conditions, times of year, times of day, and growth stages of both weed and grass.” The algorithm then efficiently differentiates between good and bad plants.³