相比于传统的强度调制/直接检测（IM/DD）系统，相干系统具有更高的容量和功率预算，能更好地满足高容量无源光网络（PON）的需求。近年来，如何将相干应用于PON场景以更好地支撑未来高带宽业务已成为研究热点。文章从系统架构、相干简化、上行突发模式检测以及灵活PON 4个方面对相干PON关键技术研究现状进行了总结，并展望提出了激光共享上下行滤波器组多载波（FBMC）-PON和半导体光放大器（SOA）电流调控灵活PON方案。

Data centers, the engines of the global Internet, rely on powerful high-speed optical interconnects. In optical fiber communication, classic direct detection captures only the intensity of the optical field, while the coherent detection counterpart utilizes both phase and polarization diversities at the expense of requiring a narrow-linewidth and high-stability local oscillator (LO). Herein, we propose and demonstrate a four-dimensional Jones-space optical field recovery (4-D JSFR) scheme without an LO. The polarization-diverse full-field receiver structure captures information encoded in the intensity and phase of both polarizations, which can be subsequently extracted digitally. To our knowledge, our proposed receiver achieves the highest electrical spectral efficiency among existing direct detection systems and potentially provides similar electrical spectral efficiency as standard intradyne coherent detection systems. The fully recovered optical field extends the transmission distance beyond the limitations imposed by fiber chromatic dispersion. Moreover, the LO-free advantage makes 4-D JSFR suitable for photonic integration, offering a spectrally efficient and cost-effective solution for massively parallel data center interconnects. Our results may contribute to the ongoing developments in the theory of optical field recovery and the potential design considerations for future high-speed optical transceivers.

Secure distribution of high-speed digital encryption/decryption keys over a classical fiber channel is strongly pursued for realizing perfect secrecy communication systems. However, it is still challenging to achieve a secret key rate in the order of tens of gigabits per second to be comparable with the bit rate of commercial fiber-optic systems. In this paper, we propose and experimentally demonstrate a novel solution for high-speed secure key distribution based on temporal steganography and private chaotic phase scrambling in the classical physical layer. The encryption key is temporally concealed into the background noise in the time domain and randomly phase scrambled bit-by-bit by a private chaotic signal, which provides two layers of enhanced security to guarantee the privacy of key distribution while providing a high secret key rate. We experimentally achieved a record classical secret key rate of 10 Gb/s with a bit error rate lower than the hard-decision forward error correction (HD-FEC) over a 40 km standard single mode fiber. The proposed solution holds great promise for achieving high-speed key distribution in the classical fiber channel by combining steganographic transmission and chaotic scrambling.

We develop and experimentally demonstrate a phase-sensitive continuous variable quantum key distribution system with improved secure key rate. This is achieved using multimode coherent states with phase-conjugated subcarrier modulation and phase-sensitive detection. The local oscillator for phase-sensitive detection is regenerated from a polarization-multiplexed carrier wave via optical injection locking. The proposed scheme has a higher classical information capacity at a given number of received photons and exhibits a higher secure key rate when applying the security analysis of the GG02 protocol. Experimental results confirm the higher secret key rate and better excess noise tolerance of the new scheme compared to the typical implementation of GG02.

Integrated optomechanical crystal (OMC) cavities provide a vital device prototype for highly efficient microwave to optical conversion in quantum information processing. In this work, we propose a novel heterogeneous OMC cavity consisting of a thin-film lithium niobate (TFLN) slab and chalcogenide (ChG) photonic crystal nanobeam coupled by a wavelength-scale mechanical waveguide. The optomechanical coupling rate of the heterogeneous OMC cavity is optimized up to 340 kHz at 1.1197 GHz. Combined with phononic band and power decomposition, 17.38% energy from the loaded RF power is converted into dominant fundamental horizontal shear mode (SH0) in the narrow LN mechanical waveguide. Based on this fraction, as a result, 3.51% power relative to the loaded RF energy is scattered into the fundamental longitudinal mode (L0) facing the TFLN-ChG heterogeneous waveguide. The acoustic breathing mode of the heterogeneous OMC is successfully excited under the driving of the propagating L0 mode in the heterogeneous waveguide, demonstrating the great potentials of the heterogeneous piezo-optomechanical transducer in high-performance photon–phonon interaction fields.

随着国家电网规模的不断扩大，架空导线作为电力系统的重要组成，对它的定期巡检变得极其重要，同时，随着低空飞行领域的开放，为了保证国家电网的正常运行及低空飞行的安全，架空导线的识别也变得极其重要。文中提出了一种使用Deeplabv3+语义分割网络模型对红外航拍图像架空导线进行识别的方法，并且针对红外架空导线图像目标的特征对该算法进行了改进。首先在原Deeplabv3+算法的特征提取主干网络ResNet50中加入注意力机制，使网络突出导线目标所在区域的特征，更加关注导线目标所在的位置，进而弱化背景等非主要区域的特征。然后对Deeplabv3+的编码器部分进行改进，在ResNet50模型中加入特征金字塔网络，可以将浅层和深层的特征进行融合，增强网络对不同大小目标属性的识别能力，及导线这种小目标的检测能力，进而提高网络的整体识别效果。实验结果表明：改进后的算法检测性能良好，均像素精度为93.52%，平均交并比为87.83%。

Scattering-induced glares hinder the detection of weak objects in various scenarios. Recent advances in wavefront shaping show one can not only enhance intensities through constructive interference but also suppress glares within a targeted region via destructive interference. However, due to the lack of a physical model and mathematical guidance, existing approaches have generally adopted a feedback-based scheme, which requires time-consuming hardware iteration. Moreover, glare suppression with up to tens of speckles was demonstrated by controlling thousands of independent elements. Here, we reported the development of a method named two-stage matrix-assisted glare suppression (TAGS), which is capable of suppressing glares at a large scale without triggering time-consuming hardware iteration. By using the TAGS, we experimentally darkened an area containing 100 speckles by controlling only 100 independent elements, achieving an average intensity of only 0.11 of the original value. It is also noticeable that the TAGS is computationally efficient, which only takes 0.35 s to retrieve the matrix and 0.11 s to synthesize the wavefront. With the same number of independent controls, further demonstrations on suppressing larger scales up to 256 speckles were also reported. We envision that the superior performance of the TAGS at a large scale can be beneficial to a variety of demanding imaging tasks under a scattering environment.