Researchers at Tsinghua University in China have developed a groundbreaking photonic chip that can process, transmit, and reconstruct images in mere nanoseconds. This innovative chip bypasses the traditional optical-to-electronic data conversion used by conventional chips, significantly enhancing the speed and efficiency of image processing.

Machine vision, a rapidly evolving field where cameras, sensors, and algorithms collaborate to interpret the world and execute specific tasks, has traditionally relied on transferring data over long distances for analysis. This process, however, is too slow for many real-time applications. “The world is entering an AI era, but AI is very time- and energy-exhaustive,” said Lu Fang, associate professor at Tsinghua University’s Department of Electronic Engineering. In today’s fast-paced world, machine vision now requires on-device data processing, known as edge computing, to enable quicker decision-making.

“The growth of edge devices, such as smartphones, intelligent cars, and laptops, has led to an explosive increase in image data that needs to be processed, transmitted, and displayed. We are working to advance machine vision by integrating sensing and computing in the optical domain,” Fang explained.

Edge tasks, like autonomous driving, are hindered by the millisecond delays caused by converting optical data to electronic formats. “Capturing, processing, and analyzing images for edge-based tasks such as autonomous driving is currently limited to millisecond-level speeds due to the necessity of optical-to-electronic conversions,” Fang noted. Presently, machine vision devices capture images in their optical form, which must be converted to electronic data for computational processing.

Under Fang’s guidance, the research team developed an Optical Parallel Computational Array (OPCA) chip with a sensing-computing array made using ring resonators. This innovative design allows the photonic chip to convert an optical image into a two-dimensional light intensity representation, which is then guided onto the chip using a micro-lens array.

The OPCA chip boasts a processing bandwidth of up to a hundred billion pixels and a response time of just six nanoseconds. By processing data as light signals, the researchers created an all-optical neural network, which they deployed for classification tasks typically carried out on edge devices.

“Because each sensing-computing element of this chip is reconfigurable, they can each operate as a programmable neuron that generates light modulation output based on the input and weight,” added Fang. “The neural network connects all the sensing-computing neurons with a single waveguide, facilitating an all-optical full connection between the input information and the output.”

The research team showcased the chip’s capabilities by deploying it for tasks such as image classification of hand-drawn images and image convolution, where a filter extracts specific features from an image. The successful completion of these tasks demonstrates the chip’s potential in handling complex image processing tasks.

The research team aims to increase the overall size of the OPCA chip and enhance the neural network’s processing capacity to bring it closer to commercial use. This revolutionary photonic chip could transform the fields of machine vision and edge computing, paving the way for faster and more efficient real-time image processing.

By integrating advanced photonic technology with edge computing, this innovation holds promise for numerous applications, from autonomous vehicles to intelligent devices, heralding a new era in AI-driven image processing.

By Impact Lab