基于特征融合与双模板嵌套更新的孪生网络跟踪算法

为提高全卷积孪生网络SiamFC在复杂场景下的识别和定位能力,提出一种基于多响应图融合与双模板嵌套更新的实时目标跟踪算法。使用深度ResNet-22替换AlexNet作为骨干网络以提升网络特征提取性能,建立强识别能力的骨干语义分支。在ResNet-22的浅层使用高分辨率特征,构造强定位能力的浅层位置分支,计算并融合两个分支响应。通过高置信度的双模板嵌套更新机制对两个分支的模板进行更新,以适应目标的外观和位置变化。在OTB2015和VOT2016数据集上的实验结果表明,与基于SiamFC、SiamDW等的目标跟踪算法相比,该算法在目标快速移动、遮挡等复杂场景下跟踪效果更稳定,并且运行速度达到34 frame/s,满足实时性要求。     

双模板剂B-EU-1/ZSM-5复合分子筛的快速合成及催化应用

以不同氢氧化物做碱源合成B-EU-1/ZSM-5复合分子筛, 并比较各平衡阳离子的结构导向作用及其形貌特征, 发现钠离子的导向作用最强, 且合成复合分子筛的表面较光滑、无明显晶面界限。对合成样品进行了TG-DTG、N₂吸附-脱附和NH₃-TPD分析。结果表明: 双模板剂一步法合成的复合分子筛模板剂脱除的失重率为5.67%, 小于机械复合分子筛的失重率7.31%, 复合分子筛的比表面积、孔容和微孔平均孔径均有所增大, 同时其酸强度、酸量都增大, 有利于二甲苯优先从复合分子筛的孔道结构中扩散出来。将分子筛应用于甲醇转化制二甲苯催化反应, 结果显示, 双模板剂一步法合成的复合分子筛催化产物油相中芳烃的选择性最高达到66.72%, 芳烃中二甲苯的含量为46.15%, 二甲苯中对二甲苯的含量达到30.75%。这是由于B-EU-1/ZSM-5催化剂的特殊孔道结构具备择形效应, 使得分子动力学直径较小的二甲苯分子优先从其孔道中扩散出来。     

片状形貌ZSM-39分子筛的合成

利用高通量水热合成装置,使用正丙胺(R1)和六亚甲基四胺(R2)双模板剂,合成片状形貌的ZSM-39分子筛。采用XRD、SEM、FTIR、TG和N2吸附-脱附等方法对试样进行了表征,考察了晶化时间和双模板剂比例对合成的ZSM-39分子筛形貌的影响。实验结果表明,当晶化时间足够长(120 h)时,所得ZSM-39分子筛的结晶度良好;合成的ZSM-39分子筛的形貌可通过调节双模板剂的比例进行有效控制,R1含量过高或过低时,均不能得到片状形貌的ZSM-39分子筛;在晶化温度180℃、晶化时间120 h、双模板剂比例n(R1)∶n(R2)=0.5的条件下,所得ZSM-39分子筛呈现均匀薄片状形貌,厚度50 nm左右。     

双模板剂法SAPO-34分子筛的合成及其MTO催化性能调变

以四乙基氢氧化铵（TEAOH）和二乙铵（DEA）为混合模板剂，在低投料硅铝比[n (SiO₂) ∶n (Al₂O₃)=0.2]及低模板剂用量[n (模板剂) ∶n (Al₂O₃)=1.9]下，考察了两种模板剂比例的调变对合成的SAPO-34分子筛物化性能及其催化甲醇制烯烃反应（MTO）催化性能的影响。研究发现，通过改变两种模板剂比例，可以明显调变SAPO-34分子筛晶粒尺寸、硅分布（晶粒表面和体相的硅分布）、硅原子的配位环境，从而影响其MTO催化反应的效果。在低模板剂用量制备的SAPO-34产品中，晶粒尺寸是影响其催化寿命的最主要因素，小晶粒分子筛因其扩散路径短有利于延长催化寿命。此外，硅分布也是影响催化寿命的因素之一，表面富硅的分子筛导致外表面积炭程度大于晶内积炭，积炭趋势由外向内发展，加速分子筛“假性”失活。硅分布还影响MTO反应产物分布，表面富硅分子筛外表面更易发生非择形催化，显著提高C₄～C₆等产物的选择性，不利于目标产物双烯（乙烯+丙烯）选择性的提高。     

Hybrid dual-template induced nitrogen-doped hierarchically porous carbon as highly efficient oxygen reduction electrocatalyst

Fuel cell, as a promising future energy device, is one of the potential electric generators maintaining the sustainable development of our society and alleviating the major problems related to the energy shortage and environment pollution. However, the high cost and the limited resource of Pt remain the key obstacles for the commercialization of fuel cells. A hybrid dual-template strategy is developed to synthesize the nitrogen-doped ordered hierarchically porous carbon (NHMC) via a surfactant-templating organic resol self-assembly with F127 as soft template and SiO₂ nanosphere as the hard-template. The NHMC catalyst presents three-dimensional hierarchically porous structure, composed of small ordered mesopore (∼3.8 nm), large 3-D interconnected mesopore (∼12 nm) as well as micropore. The catalyst exhibits tailored pore structures and a well-tuned surface chemical environment. It displays high onset potential of 0.91 V and half-wave potential of 0.76 V as well as high limiting current via four-electron pathway for oxygen reduction reaction (ORR). The NHMC also shows high stability and excellent methanol tolerance. This work brings inspiration for the synthesis of low-cos Pt-free catalyst with high activity, stability as well as high methanol tolerance.     

Dual-template synthesis of hierarchical porous carbon and its application in electrical double-layer capacitors

Hierarchical porous carbons (HPCs) with tailored pore structure were fabricated via direct carbonization of a precursor consisting of agar and acetates in an inert atmosphere. The agar served as the carbon source and curing agent, while potassium acetate and calcium acetate were used to create micropores and mesopores, respectively. The morphology and structural features of the HPCs were characterized, and the electrochemical properties were estimated in a three-electrode system with 6 M KOH as the electrolyte. The resulting HPCs possess a developed pore structure and concentrated pore size distribution (focused at 0.4 and 4 nm), and exhibit typical electrical double-layer capacitive behaviour. Specially, the HCP-2 owns a large specific surface area of 1441.65 m²/g and excellent electrochemical properties, including a favourable specific capacitance (235 F/g at 1 A/g), good rate capability (capacitance retention rate of 88.1% at 20 A/g), and impressive long-term cycling stability (90.6% retention of initial capacitance after 3000 cycles).