Researchers in China have introduced an intelligent photonic sensing-computing chip capable of processing one hundred billion pixels in just 6 nanoseconds.
This groundbreaking development promises to revolutionize high-speed image processing, offering major benefits for edge intelligence (edge AI) applications such as autonomous driving, industrial inspection, and robotic vision.
Edge AI allows the deployment of machine learning algorithms to the data-generating edge device.
Photonic Chip Boasts High-speed Processing Speeds
The photonic chip, described in the study published in Optica, is known as the optical parallel computational array (OPCA) chip. It boasts a processing bandwidth of up to one hundred billion pixels and a response time of merely 6 nanoseconds.
This speed is approximately six orders of magnitude faster than current image processing methods, which are typically limited to millisecond-level speeds due to the necessity of optical-to-electronic conversions.
HSINCHU, TAIWAN - SEPTEMBER 16: A closeup of a silicon wafer on display at Taiwan Semiconductor Research Institution on September 16, 2022 in Hsinchu, Taiwan. Photo by Annabelle Chih/Getty Images
Edge Computing's Need for Speed
Edge computing, which involves performing intensive computing tasks like image processing and analysis on local devices, is evolving into edge intelligence with the integration of artificial intelligence (AI). However, current technology faces significant challenges.
According to Lu Fang from Tsinghua University, "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."
The new OPCA chip circumvents this limitation by performing all processes in the optical domain, drastically enhancing speed and efficiency.
Traditional machine vision systems are limited by the need to convert optical data into electrical signals for processing, a process that slows down speed and limits capacity. The OPCA chip overcomes these limitations by keeping all processing within the optical domain.
This innovative chip uses an array of ring resonators to directly transform an optical image into a coherent light signal that can be processed on the chip. A micro-lens array focuses the image onto the chip, enabling the creation of a complete optical neural network.
This design allows modulated light signals to be coupled into a high-bandwidth optical waveguide, significantly improving overall efficiency and performance.
What's Next?
The OPCA chip's potential applications are vast and varied. Wei Wu, co-first author of the paper, noted the chip and optical neural network could boost the efficiency of processing complex scenes in industrial inspection and help advance intelligent robot technology to a higher level of cognitive intelligence.
The researchers demonstrated the chip's capabilities by using it to classify handwritten images and perform image convolution, indicating its potential for widespread applications.
Looking ahead, the research team is working on improving the OPCA chip to further enhance computational performance and align it more closely with real-world scenarios.
They aim to optimize the chip for edge computing applications and increase its processing capacity to handle more complex and realistic intelligent tasks. Reducing the form factor of the chip is also a priority to facilitate practical use.
Fang expressed hope that machine vision will be gradually improved to be faster and more energy-efficient by using light to perform both sensing and computing.
