一种基于双通道的水下图像增强卷积神经网络

近年来各国对于海洋生物的保护意识日益强烈,用来监测海洋生物生存状态的水下机器人装备的研发是保护海洋生物资源的关键。水下相机是这类机器人在水下进行海洋生物监测时的光学感知设备。然而水下环境复杂,拍摄到的图像模糊不清,为解决水下图像模糊等问题,提出了一种基于双通道的水下图像增强卷积神经网络。在网络的编码器中采用双通道结构,其中一个通道采用了密集连接和高效通道注意力机制,提取水下图像的细节特征,另一个通道采用多尺度结构,提取原始图像的多尺度语义特征。接着,在网络中引入残差注意力模块和自适应特征融合模块,进一步优化了特征。最后将优化后的特征输入解码器重建出增强后的水下图像。试验表明：提出的网络算法在UIQM指标和Entropy指标上分别为3.005 6和7.654 7,较第二名的算法分别高出0.097 5和0.123 2。     

2020年7月两次入海江淮气旋爆发性与非爆发性发展特征对比分析

利用高时空分辨率的欧洲中期天气预报中心(European Centre for Medium-Range Weather Forecasts, ECMWF)对全球气候的第五代大气再分析数据集(European Re-Analyses 5, ERA5),对比分析了2020年7月淮河上游地区的两次江淮气旋过程,并结合海洋—大气—波浪—泥沙耦合模式(Coupled Ocean-Atmosphere-Wave-Sediment Transport modeling system, COAWST)进行敏感性试验,探讨同一气候尺度背景下,高低频大气环流形势和海气耦合作用对江淮气旋的影响。结果表明,低空环流形势的高频变化和海表温度的升高对气旋的大风影响和强度发展有重要作用。低空环流形势中,存在“气旋—反气旋—气旋”环流天气尺度波列,易造成气旋大风叠加增强;海气交互界面的高海表温度加热作用导致的感热和潜热释放通过气旋北部弯曲锋面强烈的上升运动,为气旋发展提供能量,促进气旋入海后爆发性增强。     

CT强化幅度与食管癌病理形态学和血管内皮生长因子表达间的关系

目的:对56例食管癌患者血管内皮生长因子(VEGF)表达结果与CT强化峰值参数间进行对照,研究其相关性,为食管癌的影像学诊断及早期发现提供参考依据。方法:56例患者均于术前行CT平扫+动态增强扫描,记录不同扫描时期病灶的CT平扫值、增强峰值;VEGF在正常食管粘膜和食管癌组织中的表达用免疫组织化学法(SP法)测定。所有数据均通过SPSS 21.0统计软件处理。结果:VEGF的表达增强可能是鳞状细胞癌恶性度增高及发生淋巴结转移的标志,CT扫描增强幅度与食管癌中VEGF表达强弱无关。结论:VEGF的表达增强可能是鳞状细胞癌的发生及淋巴结发生转移的标志,CT扫描中病灶增强幅度不能反应肿瘤淋巴结的转移。      

Feasibility of signal enhancement with multiple grounded‐wire sources for a frequency‐domain electromagnetic survey

Frequency‐domain electromagnetic methods with a grounded‐wire source are powerful tools in geophysical exploration. However, the signal may be too weak to guarantee the quality of survey data in complex electromagnetic environments, especially when the receiver is located in the air for the newly developed grounded‐source airborne frequency‐domain electromagnetic method. In this paper, a signal enhancement method with multiple sources is proposed to solve this problem. To evaluate the signal enhancement effect, we compared the signals generated by a single source and multiple sources with equal electric moment. The signal differences caused by synchronisation error and separation distance between source elements were analysed, and the methods to achieve maximum signal were introduced. Besides, we discussed the interaction between adjacent source elements to ensure the system safety, including the changes in output current and the safe distance between two sources using a dual‐source model. Lastly, a comprehensive field experiment was designed and conducted to test the multiple‐source method. The data processing results are comparable for single and dual sources, and the signal‐to‐noise ratio of dual source is higher in the field test. The subsurface resistivity structure at the test site is consistent with the previous controlled‐source audio‐frequency magnetotellurics method. These results show that signal enhancement with multiple sources is feasible. This study provides guidance to the application of multiple sources in field surveys when the survey environment is complex and rigorous.     

2009年夏季极区电离层加热实验观测研究

2009年夏季,中国在挪威特罗姆瑟(Tromsφ)利用非相干散射雷达进行电离层加热实验。实验中发现了处在150km附近的小范围温度增强事件和150—400km之间的大范围温度增强事件。前者的相对增幅明显,后者的绝对增量更大。对这两类加热事件来说,温度增量百分比均随加热功率的增加线性增长,但对比两类加热事件来看,增长速度随加热频率增大而减小。加热效应存在显著的二维分布特征,加热最强方向处在场向附近。加热效应受加热天线波束指向与地磁场的夹角影响明显,随夹角增大温度增强效应减弱。     

一种边缘特征增强的多源遥感影像融合方法研究

基于分频分策略的融合思想,提出了一种边缘特征增强的多源遥感影像融合方法。实验证明,融合后的影像不仅在很大程度上保留了多光谱影像的光谱特征,还提高了空间分辨率,特别是边缘特征得到进一步增强。     

探讨图像增强中直方图均衡化的应用

图像增强是数字图像的预处理,能有效地改善图像整体或局部特征。主要研究了直方图均衡化的图像增强的基本原理,并给出了相关推导公式和算法,同时以一个灰度图像为例,在VC++集成开发环境下实现了直方图均衡化的增强处理,并给出了实验结果图像及直方图。     

包头市昆都仑水库增雨燃烧炉首次作业的效果分析

通过2009年5月27日包头市昆都仑水库地面增雨燃烧炉作业实况和多普勒天气雷达速度、713天气雷达回波及区域地面自动雨量站资料的对比分析,发现在深厚的层状云系中,适时进行地面燃烧炉增雨作业,可增强区域降水量,提高增雨作业能力。分析结果为指导包头地区地面燃烧炉布点和增雨作业提供了科学依据。     

Application of a cascading filter implemented using morphological filtering and time–frequency peak filtering for seismic signal enhancement

Inspired by the idea of the iterative time–frequency peak filtering, which applies time–frequency peak filtering several times to improve the ability of random noise reduction, this article proposes a new cascading filter implemented using mathematic morphological filtering and the time–frequency peak filtering, which we call here morphological time–frequency peak filtering for convenience. This new method will be used mainly for seismic signal enhancement and random noise reduction in which the advantages of the morphological algorithm in processing nonlinear signals and the time–frequency peak filtering in processing nonstationary signals are utilized. Structurally, the scheme of the proposed method adopts mathematic morphological operation to first preprocess the signal and then applies the time–frequency peak filtering method to ultimately extract the valid signal. Through experiments on synthetic seismic signals and field seismic data, this paper demonstrates that the morphological time–frequency peak filtering method is superior to the time–frequency peak filtering method and its iterative form in terms of valid signal enhancement and random noise reduction.