A joint research team from the Korea Advanced Institute of Science and Technology (KAIST) and the Gwangju Institute of Science and Technology (GIST) has developed a next-generation image sensor that adapts to extreme changes in brightness without requiring external correction. The technology, detailed in the journal Advanced Materials, represents a potentially important step forward for autonomous vehicles, robotics, and advanced security systems.
Tripartite Synapse-Inspired In-Sensor Processing
The new device is a ferroelectric-gated phototransistor (FGPT) that integrates sensing and processing directly within the sensor. Inspired by a tripartite synapse architecture found in the brain, the sensor simultaneously performs light detection, recording, and neuromorphic processing. This eliminates the need for traditional post-processing hardware or software that conventional image sensors rely on when handling dramatic shifts in exposure.
By using partial polarization switching in the ferroelectric gate insulator, combined with photogating effects, the device can linearly control photoactive channels and store image information as non-volatile memory. This design enables real-time adaptation to bright or dim conditions without data loss, reducing the likelihood of overexposure or missed detail in high-contrast environments.
Performance and Applications
Tests showed the device achieves a dynamic range of 153 dB, far exceeding the performance of conventional sensors. The integrated in-sensor processing also delivered about a 40 percent improvement in face recognition accuracy under variable lighting conditions. These capabilities make it particularly suitable for real-world applications where lighting is unpredictable, such as autonomous driving, biometric security, and robotic vision systems.
Professor Song Young-min of KAIST and Professor Kang Dong-ho of GIST led the study, with support from the National Research Foundation of Korea. According to the foundation, the device “can implement light detection, recording, and processing within the device itself, leading to the development of a next-generation image sensor.”
Broader Research Context
This innovation builds on KAIST’s broader research into bio-inspired sensing technologies, including earlier acoustic sensor developments modeled on the human ear. It also reflects growing international interest in ferroelectric neuromorphic devices, which combine low power consumption with the ability to mimic neural processing. As demand for high-performance visual AI systems expands, the FGPT highlights how biomimetic approaches can overcome long-standing challenges in imaging technology.
Sources: Chosun Biz, KAIST News, KAIST
–
By Ali Nassar-Smith
