Streamlined computer vision can map indoors

Written by: Andrew Wade | Published:
(Credit: University of Bristol)

A new computer vision platform that uses an evolving pixel map to localise itself could have applications ranging from robotics to autonomous vehicles.

Developed at the University of Bristol, the low-energy system is a sensor-processor called a Pixel Processor Array (PPA), with the technique itself known as Weighted Node Mapping. As images are sensed, the processor decides which pixels are required for the task at hand, keeping only those deemed necessary for localisation and mapping. This instant mapping functionality does not rely on energy intensive technologies such as GPS and also works just as well indoors as outside, giving it significant potential as a low-cost, efficient computer vision system of the future.

“We often take for granted things like our impressive spatial abilities,” said team leader Walterio Mayol-Cuevas, Professor in Robotics, Computer Vision and Mobile Systems at the University of Bristol’s Department of Computer Science. “Take bees or ants as an example. They have been shown to be able to use visual information to move around and achieve highly complex navigation, all without GPS or much energy consumption.

“In great part this is because their visual systems are extremely efficient and well-tuned to making and using maps, and robots can't compete there yet.”

The research formed part of the MSc dissertation of Bristol robotics student Hector Castillo-Elizalde, who helped design the algorithm that runs on the PPA. According to the team, the algorithm is deceptively simple, deciding if each new frame is sufficiently different from the last, then retaining whatever data is judged to be useful.

As the device moves around an area – on a robot or carried by a human - it builds a visual catalogue that can then be used to match any new image when the system is seeking to localise itself. However, no full images leave the PPA, only the key data points that indicate where it is with respect to the visual catalogue. According to researchers, this not only makes the system more energy efficient, it also provides a layer of privacy that other computer visions are lacking. A research paper based on the work has been accepted for publication in the IEEE International Conference on Robotics and Automation (ICRA) 2021.

This material is protected by MA Business copyright see Terms and Conditions.
One-off usage is permitted but bulk copying is not. For multiple copies contact the sales team.


Please view our Terms and Conditions before leaving a comment.

Change the CAPTCHA codeSpeak the CAPTCHA code