一种基于FCOS神经网络的小建筑物目标检测方法北大核心

提出了一种基于FCOS神经网络的小建筑物目标检测算法,针对FCOS模型在特征提取阶段提取到的小建筑物目标特征较少问题,引入多尺度检测和可变形卷积方式,加强网络对小建筑物目标的特征提取能力,并通过改进后的SGE注意力机制降低特征图中的干扰噪声权重。改进后的网络可以提取到更多的小建筑物目标特征,对环境干扰噪声的鲁棒性更强。在自己搭建的数据集上进行了实验测试,结果表明,在相同环境下网络改进后建筑物的整体检测准确率提升了1.7%,其中对小建筑物目标提升了3.6%,减少了小建筑物目标漏检、误检的问题。     

Computational and experimental study of dental resin composites with high filler content

Journal of Materials Science - Nanocomposites have been used in various industries due to their excellent properties. For some of them, high filler content (over 50 wt%) is beneficial to improving...     

立交桥现浇箱梁预应力张拉及管道真空压浆技术

依托长山大道混凝土箱梁施工，对现浇箱梁预应力伸长量和张拉力进行计算，重点介绍了张拉程控原理、管道真空压浆工艺及操作要点，保证了预应力的有效施加，管道压浆饱满密实，提高了工效，保障了施工安全，可为类似工程提供借鉴。     

Stacked Pyramid Attention Network for Object Detection

Neural Processing Letters - Scale variation is one of the primary challenges in object detection. Recently, different strategies have been introduced to address this challenge, achieving promising...     

盾构隧道施工物料无人运输发展现状及关键技术

盾构隧道施工物料无人运输有助于提升运输效率、降低运输成本、减少运输安全事故率。与公路汽车无人驾驶、港口车辆无人驾驶、轨道交通无人驾驶等逐步成熟的无人驾驶技术不同，在隧道场景下实施施工物料无人运输存在运输物料种类繁多、运输调度困难、地下定位信号拒止、狭窄车道频繁会车、行车路面工况复杂、地上地下联动响应慢等诸多挑战。从实现盾构隧道施工物料无人运输的关键问题分析入手，综述当前盾构隧道施工物料运输方式、运输需求、无人运输发展现状及存在的挑战，提出盾构隧道施工物料无人运输的五大关键技术：多种类物料智能货控能效管理技术，无人化垂直装卸门机控制技术，多传感器融合同步定位与地图构建技术，隧道复杂环境路径规划与自主避障技术，隧-地一体化联动响应的高可靠性无线通信技术。     

Improved thermal stability of the near-infrared Al-modulated Zn3Ga2GeO8: Cr3+ phosphors for plant growth applications

The exploration of the high thermal stability near-infrared (NIR) phosphors is significantly crucial for the development of plant lighting. However, NIR phosphors suffer from the poor chemical and thermal stability, which severely limits their long-term operation. Here, the successful improvement of luminous intensity (149.5%) and thermal stability at 423 K of Zn₃Ga₂GeO₈ (ZGGO): Cr³⁺ phosphors is achieved for the introduction of Al³⁺ ions into the host. The release of carriers in deep traps inhibits the emission loss for the thermal disturbance. Furthermore, an NIR light emitting diodes (LEDs) lamp is explored by combining the optimized Zn₃Ga_1.1675Al_0.8GeO₈: 0.0325Cr³⁺ phosphors with a commercial 460 nm blue chip, and the emission band can match well with the absorption bands of photosynthetic pigments and the phytochrome (P_R and P_FR) of plants. The explored LEDs lamp further determines the growth and the pheromone content of the involved plants for the participation of the NIR emission originated from Cr³⁺ ions. Our work provides a promising NIR lamp as plant light with improved thermal stability for long-term operation.