#Finland #Espoo #neuromorphic computing #MISEL Project #VTT Technical Research Centre

Energy-Efficient Machine Vision Inspired by Human Eyesight

In a remarkable breakthrough led by the VTT Technical Research Centre of Finland, European researchers have successfully developed a machine vision system inspired by human eyesight. This innovative project, known as the MISEL (Multispectral Intelligent Vision System with Embedded Low-Power Neural Computing), utilizes neuromorphic computing to mimic the cooperation seen between the eye and the nervous system while being deployed on edge-computing circuits.

Initiated in 2021, the MISEL project is nearing its completion after substantial backing of nearly EUR 5 million from the EU's Horizon 2020 programme. The integration of neuromorphic computing—which replicates the brain's information processing capabilities—with semiconductor technologies positions this project at the forefront of intelligent device development.

Pioneering Smart Devices for Independence

According to Jacek Flak, the Research Team Leader at VTT, the main aim is to create truly smart devices capable of making independent observations and decisions without relying on data transmission to cloud-based supercomputers. This is crucial as it can drastically reduce the energy required compared to traditional digital processing methods, which may consume hundreds to thousands of times more energy.

Flak highlights the significance of edge computing — a process that occurs at or near the source of data generation, which significantly enhances the efficiency of devices such as drones and robots. For instance, these devices can perform rescue operations post-earthquakes without needing constant network connections or large battery packs.

Nature as an Inspiration

In designing this system, the project drew inspiration from nature, particularly the complex interplay between the human retina and brain areas responsible for visual interpretation. Additionally, the fruit fly served as a compelling model for efficiency in energy consumption while navigating its environment.

One of the key deliverables of the MISEL project is a specialized system-on-chip brought forth by Kovilta, a company that specializes in advanced integrated circuits. This innovative chip combines both imaging with significant image processing capabilities, resulting in an ability to detect motion and changes over time, reflecting how a biological eye operates.

Advanced Capabilities and Low Power Consumption

The machine vision technology developed under this project boasts features such as high dynamic range image sensing (with a capability exceeding 120 dB), high frame rates (over 1000 frames per second), and parallel image processing, allowing for sophisticated motion analysis and pattern recognition. Unlike traditional video cameras, that capture still frames, this new technology follows dynamic scenes, enabling a more accurate analysis of the environment.

Moreover, researchers explored the potential of quantum dot image sensors. This advanced camera technology expands visibility into the infrared range, enhancing the machine's ability to detect movement in challenging conditions such as low light or fog.

Unified System Design for Enhanced Efficiency

An essential aspect of MISEL was the co-design of sensors, memory units, algorithms, and electronic components as part of a unified system. This strategy maximizes efficiency while limiting overall energy consumption. Through the collaborative efforts with Lund University, specialized non-volatile memories were developed, which can be directly integrated onto chips, further enhancing system efficiency.

The goal is to deploy these accelerated architectures in fields such as autonomous robotics and vehicle technology to improve autonomous operational capabilities. Mika Laiho, the Chief Technology Officer at Kovilta, emphasizes that understanding surroundings and making quick decisions is paramount for robots and vehicles to function safely and independently among humans.

Broad Applicability Across Industries

The implications of these developments extend well beyond research, presenting opportunities for various applications, including smart cameras for monitoring industrial environments, mobile robots making autonomous decisions, and much more. The next steps will involve leveraging the findings from MISEL in upcoming projects and pilot production lines at VTT.

Ultimately, the vision is clear: to develop autonomous devices that are capable of seeing, thinking, and acting as independently and energetically as a fruit fly, revolutionizing the future of machine vision technology.

For further inquiries, please contact:

• - Jacek Flak, Research Team Leader, MISEL Project Coordinator, VTT Technical Research Centre of Finland.
• - Email: [email protected]

Join us as we move toward a future where machines can independently and intelligently interact with their environments in a sustainable manner.