Recently, a team from the School of Physics at Southeast University, including Prof. Zhenhua Ni, Prof. Junpeng Lu, Associate Researcher Wenhui Wang, and Postdoctoral Researcher Kaiyang Liu, has made progress in non-pixelated image sensing. The work, titled “Non-pixelated in-materia retinomorphic sensor via photocarrier dynamics for precise spatiotemporal perception,” has been published in Nature Communications.
With the rapid development of intelligent sensing technologies in fields such as robotics, virtual reality, sports science, and smart monitoring, higher demands are being placed on the real-time acquisition and interpretation of dynamic visual information. In real-world scenarios, visual data inherently contain both spatial structure and temporal variation—referred to as spatiotemporal information. However, conventional CMOS image sensors operate with independently functioning pixels, making it difficult to directly extract such spatiotemporal correlations at the sensing stage. Instead, they rely on high-resolution imaging and complex backend algorithms for detection and recognition.This traditional “imaging-then-computing” paradigm leads to redundant data transmission, increased power consumption, and higher latency, which has become a key bottleneck for achieving high-efficiency, low-latency perception in edge intelligent systems.
To address this challenge, the joint team from Southeast University (Prof. Zhenhua Ni, Prof. Junpeng Lu, and Associate Researcher Wenhui Wang) and the Institute for Brain-Inspired Intelligence at Nanjing University (Prof. Feng Miao and Prof. Shijun Liang) proposed and demonstrated, for the first time, a retinomorphic sensor based on the lateral photoelectric effect. This work establishes a new in-materia, non-pixelated approach for spatiotemporal information sensing.The device leverages the diffusion and drift of photogenerated carriers to naturally encode spatial and temporal information within the material itself, thereby significantly reducing data transmission and computational load. When combined with a lightweight neural network, the system achieves high-accuracy recognition of human motion sequences. This provides a promising solution for low-power autonomous perception and practical applications of neuromorphic vision devices.
Postdoctoral Researcher Kaiyang Liu (Southeast University) and Assistant Professor Pengfei Wang (Nanjing University) are co-first authors of this work. Associate Researcher Wenhui Wang (Southeast University), Prof. Shijun Liang, and Prof. Feng Miao (Nanjing University) are co-corresponding authors. Southeast University is the first affiliated institution.This work was supported by the National Natural Science Foundation of China, the Natural Science Foundation of Jiangsu Province, and the Key Laboratory of Quantum Materials and Information Devices at Southeast University.
Link: https://www.nature.com/articles/s41467-026-72104-5
