Search

The emerging optoelectronic neuromorphic devices are widely concerned due to their capability to integrate the functions of signal sensing, memory, and processing. Although significant advancements have been made in the study of individual optoelectronic synaptic devices, the development of hardware‐level image recognition systems based on photo‐synapse arrays remains a challenge. In this study, a crosstalk‐free, easy‐to‐integrate, and scalable 8 × 8 crossbar array for optical image sensing and storage is demonstrated using vertical two‐terminal ZnO photo‐synapses with the self‐denoising function. By designing peripheral circuits, a complete hardware‐level artificial visual system is constructed that successfully implements the real‐time pattern recognition tasks for 8 × 8 pixel images. The excellent performance of the photo‐synapse array shows its remarkable ability in highly efficient optic neuromorphic computing. Additionally, an in‐sensor reservoir computing (RC) system is constructed for image recognition of handwritten digits. The system achieves a high classification accuracy of 95.1%. [ABSTRACT FROM AUTHOR]

Neuromorphic computing inspired by memristors has gained considerable attention due to its low power and easy integration. However, state‐of‐the‐art two‐terminal resistive switching memristors based on conductive filament formation suffer from high variability and poor controllability. As a three‐terminal device operated through electrochemistry and dynamic insertion/extraction of ions, the electrochemical ion synapse demonstrates deterministic control of electron conductivity based on ion doping. But, integrating the electrochemical ion synapse into crossbar arrays will pose higher challenges and lower integration density. Herein, inspired by first‐principles calculations, a two‐terminal bidirectional plasticity electrochemical artificial synapse with integrated lithium polymer electrolyte and polycrystalline tungsten oxide layer is reported. The linearity and stability of the device weight update are greatly improved by adjusting the defect concentration of the polycrystalline WO3 layer. Even after 16 000 write‐read events in the air, its performance remained almost unchanged. Moreover, it has an over‐limit protection mechanism under one‐way stimulation that exceeds the normal range. Based on this excellent stability, the authors designed and successfully simulated the "muscle memory" that the programmatical organization of the nervous system leads to proficiency in specific actions. [ABSTRACT FROM AUTHOR]

Organic artificial synapses are becoming the most desirable format for neuromorphic computing due to their highly tunable resistive states. However, repressively low analog switching range, inferior memory retention, and operational instability greatly hinder the further development of organic synapses. Herein, two donor‐acceptor copolymers consisting of electron‐deficient isoindigo coupled with variable donating moieties for three‐terminal organic synaptic transistors (TOSTs) are reported. It is found that the synaptic function and device stability of TOSTs are significantly improved by enhancing the electron‐donating strength of donor units. Polymer alkylated isoindigo‐bis‐ethylenedioxythiophene exhibits high analog switching range of 170 ×, two orders of magnitude higher than that of normal organic neuromorphic devices. They also demonstrate excellent memory retention of over 5 × 103 s, low switching energy of 13 fJ, and ultrahigh operational stability with 99% of its original current after 100 000 write‐read events in air. Furthermore, the high viability of strong donor strategy is showcased by demonstrating flexible TOSTs with stable synaptic function after repeated mechanical bending as well as organic synapses capable of simulating image information processing. Overall, this work highlights the advantages of the strong donor functionalization strategy to boost the synaptic performance and device stability of TOSTs. [ABSTRACT FROM AUTHOR]

The ability to achieve precise underwater 3D imaging is an urgent need in marine research and development. Yet, due to the strong scattering by the water, the laser beam is easily spread out in the underwater propagation, which hampers the imaging quality with low resolution. In this paper, we demonstrate a single-photon LiDAR with Bessel beam illumination, which central spot shows a stable diameter of about 0.73 mm after propagating 18 m in water. Together with a single-photon detector to sense the weak returning laser echoes, the transverse resolution reaches about a few millimeters at 17.5 m in water. In the experiment, an underwater optical image of 103.1 mm × 192.1 mm with 516 × 97 pixels is reconstructed. The result shows that our single-photon LiDAR with Bessel beam illumination greatly improves the imaging resolution in the long distance underwater sensing. [ABSTRACT FROM AUTHOR]

• The signal-to-noise ratio of the 3D point cloud was increased from -1.72 dB to 5.91 dB by self-gating method. • Solving the range ambiguity in the imaging process by employing multi-repetition rate laser pulse. • Suitable for long-range 3D imaging with high generation rate of point cloud. We demonstrate a self-gating single-photon imaging LiDAR with dual-repetition rate laser pulses at 1550 nm. In the experiment, two laser pulses with different repetition rates are used to enlarge the unambiguous range. A GHz-gated Geiger-mode InGaAs/InP single-photon detector is employed to detect high rate echo photons, and two galvanometer mirrors are used to realize laser beam scanning imaging. The self-gating denoising method is proposed to remove the randomly distributed noise in the three-dimensional point cloud, which increase the signal-to-noise ratio of the system from -1.72 to 5.91 dB. We attained a large range single-photon image with the valid point cloud generation rate of 1.3 × 105 points per second containing targets from 200 m to 1200 m, which is cutting-edge in the field of single-photon LiDAR. The clearly-reconstructed image indicates the self-gating single-photon imaging provides a promising approach for long-distance and high-speed single-photon imaging. [ABSTRACT FROM AUTHOR]

• High-speed airborne single-photon LiDAR at 1550 nm. • Quasi-continuous Geiger-mode single-photon detector with sinusoidal gating waves at 1-GHz. • Fast rough data distillation from large background noise. • Suitable for terrain mapping and 3D city modeling. We report a high-speed airborne single-photon counting LiDAR at 1550 nm. By using a quasi-continuous Geiger-mode single-photon detector with sinusoidal gating waves at 1-GHz with a 1-MHz repetition rate pulsed laser as the emitting source, a point cloud rate of 1.4 × 106 points per second was obtained at the average flight height of 620 m, while the receiving aperture of the LiDAR was only 15.3 mm. To preclude the background data in the daylight which was about 30 times higher than the signal data, we proposed a processing algorithm based on the difference of signal and noise in density distribution to distill the three-dimension information of the targets. The clear reconstructed image indicates the airborne LiDAR and its data processing algorithm were qualified for terrain mapping and three-dimensional city modeling. [ABSTRACT FROM AUTHOR]

• Single-photon Lidar for canopy detection at 1064 nm. • Multi-channel single-photon detector reducing the equivalent deadtime of the Lidar. • Multi-channel coincidence counting technique for initial removal of background noise. • Suitable for airborne Lidar imaging vegetation with high depth-resolution. We reported a miniaturized near-infrared single-photon Lidar at 1064 nm, in which a fiber-coupled Si-APD single-photon detector array with 6 elements was applied to detect the reflected photons simultaneously. By utilizing the multi-channel single-photon detector for the echo photons detection, the equivalent deadtime of the Lidar could be considerably decreased from 56 ns to 9.5 ns and the effective point cloud rate was improved from 2.48 × 104 to 1.51 × 105 points per second, both of which were about 6 times optimized than the single-channel detector system. The Lidar exhibits the striking multi-layer imaging capability to obtain a dense point cloud image of the multi-layer targets and restore its spatial structure, which is very attractive for the airborne altimeter imaging vegetation. [ABSTRACT FROM AUTHOR]

Reliable laser imaging is crucial to the autonomous driving. In unfavorable weather condition, however, it always suffers from the acute background noise and signal attenuation due to the harmful strong scattering. We demonstrate a noise-tolerant LiDAR with the help of Bessel beam illumination and single-photon detection. After a 31.5-m propagation in thick fog, the Bessel beam employed by our noise-tolerant LiDAR still owns a central spot with the diameter of 1.86 mm, which supports a receiving field of view as small as 60 µrad and a great suppression of the background noise. This noise-tolerant LiDAR simultaneously performs well both in depth and intensity imaging in unfavorable weather, which can be functioned as a reliable imaging sensor in automatic driving.