激光致声技术可以将空气中的激光信号转化为水下的水声信号，将两种最佳信道高效地结合起来，从而解决空中-水下跨界通信问题。为了提高激光致声跨界通信的保密性和光声转化效率，采用了数值计算方法对不同条件下，激光在海面三维散射特性和透射特性进行了研究。首先，为了提高仿真计算的准确性，建立了三维激光海面散射场的数学模型。其次，采用数值计算方法求解三维激光海面散射的表面积分方程，并根据入射光波的特点，对入射界面进行强、弱区划分，以提高计算效率。最后，通过室内模拟试验对仿真结果进行了验证。研究结果表明，入射角度对于汽化激光致声通信的保密性和效率有着重要的影响，为后续的系统设计和相关应用研究提供了有意义的参考。

Flexible devices provide advantages such as conformability, portability, and low cost. Paper-based electronics offers a number of advantages for many applications. It is lightweight, inexpensive, and biodegradable, making it an ideal choice for disposable electronics. In this work, we propose a novel configuration of photodetectors using paper as flexible substrates and amorphous Ga2O3 as the active materials, respectively. The photoresponse characteristics are investigated systematically. A decent responsivity yield and a specific detectivity of up to 66 mA/W and 3×1012 Jones were obtained at a low operating voltage of 10 V. The experiments also demonstrate that neither the twisting nor bending deformation can bring obvious performance degradation to the device. This work presents a candidate strategy for the application of conventional paper substrates to low-cost flexible solar-blind photodetectors, showing the potential of being integrated with other materials to create interactive flexible circuits.

Ga2O3-based avalanche photodetectors (APDs) have gained increasing attention because of their excellent photoelectric conversion capability in the UV solar-blind region. Integrating high-quality epitaxial Ga2O3 with p-type semiconductor remains an open challenge associated with the integration difficulty on alleviating its defects and dislocations. Herein, we construct an APD consisting of epitaxial β-Ga2O3/La_0.8Ca_0.2MnO₃ heterostructure. The pn junction APDs exhibit a high responsivity of 568 A/W as well as an enhanced avalanche gain of up to 3.0×105 at a reverse bias voltage of 37.9 V. The integration capability demonstrated in this work provides exciting opportunities for further development of high-performance Ga2O3-based electronics and optoelectronics.

深度学习技术使目标跟踪的精度和鲁棒性得到了很大提高，基于孪生网络的跟踪方法通过在大规模数据集上进行训练，使模型能应对目标的各种形变，缺点是无法排除相似目标的干扰。为此，提出了一种基于孪生网络的两阶段目标跟踪方法。首先，采用修改后的残差网络提取性能更优的深度特征。区域建议网络通过相关滤波调制自适应更新模板，结合时域信息过滤掉易区分的负样本；然后，通过感兴趣池化层提取候选区域固定尺度的特征，并馈送到验证网络进行更精细的分类与回归。为了提升网络对高难度样本的区分能力，采用正负样本对联合训练的方式提高特征匹配的性能。在OTB100、VOT标准测试集和UAV123无人机航拍数据集上进行了评测，实验结果表明：所提方法能明显改进基准算法的性能。