A new miniature sensor mimics human eye adaptation to light, enhancing machine vision in robotics and drones, with promising accuracy in varied lighting.
Researchers have unveiled a novel miniaturized sensor that simulates the human eye’s capability to adapt to varying light conditions, as reported in a study in Nature Communications. This sensor, known as a photomemristor, is poised to enhance the performance of machine vision applications across various platforms including robots, drones, and cameras.
The sensor addresses a critical challenge that arises in machine vision when there are abrupt shifts in brightness, a common scenario for drones and robots that must identify objects in low-light settings while also being able to recognize bright light sources such as headlights.
This innovation is part of the neuromorphic machine vision field, where sensors not only capture images but also process signals, aiming to alleviate the computational burden and enhance response times to rapid changes in visual input.
Unlike traditional event-based cameras that record changes in brightness at individual pixels, the new photomemristor offers a lightweight alternative. Event-based systems can suffer from limitations and require complex algorithms, which this new sensor attempts to bypass.
The photomemristor prototype measures approximately 0.5 mm and leverages materials like TiO2 and PEDOT:PSS. Its functionality hinges on its ability to respond to humidity levels; it becomes more conductive in low light as it absorbs moisture and less conductive in bright light when moisture evaporates.
In experimental tests, a 4 × 4 array of these sensors, coupled with an artificial neural network, was used to recognize letter patterns in environments with varying brightness. The system achieved a recognition accuracy of 91.3% and processed the information in just 7.5 seconds, demonstrating its potential efficacy in real-world applications.
