南水北调中线总干渠藻类的生态调度

南水北调中线总干渠无在线调蓄水库,对藻类生态调度过程中出现的问题开展生态调度实现策略和实施方式研究。主要实现策略包括:划定自身的调蓄区,隔离生态调度对下游的影响;采用高效的渠池运行方式,减少生态调度时蓄量的反复调整;综合考虑安全、快速、平稳等需求,设定生态调度实施进程和方式。具体实施方式包括:将总干渠划分为流速调控区、调蓄区和正常运行区,分别实施等体积、控制蓄量和闸前常水位方式运行;将生态调度过程划分为充水阶段和泄水阶段,基于流速调控目标值、持续时长和水位降幅约束条件,确定各阶段时长和各分区的闸门群调控方案等。基于2018年3月输水工况,采用明渠一维非恒定流模型,仿真总干渠上游15个渠池的藻类生态调度过程。结果表明,生态调度可在3.5 d内完成,各渠池的平均流速由0.48 m/s增至0.93 m/s,持续时间超过2 h。在整个生态调度过程中,水位变化平稳,水位变幅符合安全阈值要求,下游渠道的正常运行未受生态调度明显影响。     

日本の寒地，北海道における2030年代の水稲生育への温暖化の影響予測とその対応(日语原文)

北海道では1980年代から2010年代まで，水稲圃場栽培期間である5–9月の気温は年代とともに上昇した。そこで，直近の2010年代（2010—2019年平均）と比べて， 2つの2030年代の予測気象から，既報の関係式より水稲生育を予測した。その結果，2030年代では2010年代に比べ，限界移植日（移植早限）が水稲栽培17地域の平均で8～9日早い。また，早限出穂期が1～5日早く，晩限出穂期が1～5日遅く，安全出穂期間が2～10日長い。出穂期は1～3日早い。出穂期から晩限出穂まで2～9日長いため，遅延型冷害の発生がやや少ない。生育期別気象は，出穂前24日以降30日間では生育が早いため平均気温が同じかやや低い。出穂前10日以降40日間および出穂期以降40日間では平均気温がやや高く，日射量はやや少ない。そのため，玄米収量は96～98％とやや低く，潜在収量性を示す気候登熟量示数は同じである。障害不稔発生に関係する穂ばらみ期冷害危険期の平均気温はわずかに低いかほぼ同じであるため，冷害発生の危険性は残る。一方，不稔発生をもたらす低温域の出現頻度には，地域間で差異がある。精米蛋白質含有率は同じであるが，アミロース含有率はやや低く，やや良食味である。米粒外観品質では被害粒歩合と着色粒歩合は一定の傾向がなく，未熟粒歩合はやや高い。精米白度は同じであるが，玄米白度はやや高い。以上の予測に対する技術的対応方向を示した。     

带权决策表的属性约简

属性约简是粗糙集理论的重要应用。考虑将决策表中的每行都作为一条决策规则时，若把表中出现相同决策规则的次数作为权，可得到带权决策表。提出了关于带权决策表的正域约简相应的辨识矩阵并给出了证明，从而得到了约简算法。相比于决策表中的正域约简时发现，通过将决策表转化为带权决策表后，再利用算法1进行约简时，其在一定程度上优于前者。提出了近似分类精度约简相应的辨识矩阵并给出了证明。对于2个算法，在选取的UCI数据集上进行了实验验证。通过实验进一步说明了所提出算法的可行性和有效性。     

结合注意力的3D卷积网络脑胶质瘤分割算法

为了提升脑胶质瘤分割精度，提出一种结合注意力机制的3D卷积神经网络算法。输入3个不同尺度的图像块，经过9个卷积层和1个分类层后得到3个不同的分类结果，将分类结果与注意力学习到的权重相乘并逐体素相加得到输出。此外该算法采用了一种混合Dice损失函数与Focal损失函数的超参数损失函数。实验表明，该算法的Dice系数在整体区域、核心区域以及增强区域分别达到了95.31%、80.12%、82.25%。与已有的一种脑胶质瘤分割算法deepmedic相比，整体区域、核心区域以及增强区域的Dice系数分别提升了3%、2%、6%。在脑胶质瘤分割方面，具有重要的临床意义。     

基于卷积神经网络的单目深度估计

针对利用深度学习方法对街道图像进行深度估计，提出采用语义分割的方法解决深度图出现边界模糊等问题；估计深度通过左右视角图生成视差图进行无监督的训练。在网络模型中添加语义分割层，采取多个空洞卷积并行的结构增加感受野，同时减少了图像下采样的次数，降低了由于下采样带来的信息损失，使得的结果更加准确。这也是在深度估计中首次与空洞卷积相结合增加准确率。通过对KITTI街道数据集进行训练，与现有结果相比，除了增加检测准确性，降低错误率之外，使得效果图中的物体更加清晰，并且在效果图中还保留了一些原模型中被忽视掉的细节信息，将原始图像更加完整的表现出来。     

On optical flow techniques applied to breaking surges

Surface wave breaking is a challenging two-phase flow process which plays an important role in numerous physical processes. A highly-turbulent unsteady breaking surge was investigated experimentally in a large facility, and substantial aeration occurred in the roller. The application of three optical flow techniques (Lucas-Kanade, Horn-Schunck and Farnback) to the air-water region was tested. The results indicated that the Farnback technique provided most accurate results, although some misleading results could be obtained near the air-water boundaries of the roller. The bore generation by a rapid gate closure showed a highly-unsteady complicated velocity field, with substantial free-surface deformations, wave breaking and formation of large coherent structures before the surge detached from the gate. Further upstream, the surge propagated as a hydraulic jump in translation and the data showed a marked shear region with a recirculation zone above, showing air-water flow features comparable to stationary hydraulic jumps. The upper and lower bounds of air-water flow region yielded data implying an air-to-water velocity ratio about 4–5 for a Froude number Fr₁ = 2.1.     

Governance choices and policy outcomes in the Latin American and caribbean electricity sector

The paper first provides, for Latin America and the Caribbean, country-specific synthetic quantitative measures of the degree of adoption of governance reforms in the electricity sector, focusing on four key dimensions (market structure, private sector participation (PSP), regulatory autonomy and operational organization). These measures are then correlated with standard policy performance outcomes measures. The findings suggest that, the reforms, considered jointly, were statistically significantly associated with higher technical and process quality but not with social or service improvements. Considered individually, increased regulatory autonomy enjoyed the highest positive correlation with desirable outcomes, but PSP has no significant correlation with any of the outcomes. In sum, as of 2018, governance reforms, considered jointly, had still not delivered on some promised payoffs, notably in addressing energy poverty, although the sensitivity analysis suggests that, for some indicators, the relationships are stronger for larger than for smaller countries.     

两相区配分时间对IQ&P钢组织与性能的影响

以含Cu低碳钢为研究对象,利用SEM、EPMA和拉伸试验研究了两相区配分时间对其组织演变、元素配分以及经IQ&P处理后力学性能的影响,并利用Dictra软件对元素配分行为进行了动力学计算。结果表明,IQ处理后试验钢中的块状马氏体形成于原铁素体区域;随两相区配分时间延长,粒状马氏体数量减少,板条状马氏体之间的间距减小。检测和计算结果的对比显示,C、Mn、Cu 3种元素的相对配分速率与计算结果一致,但实际配分速率低于各自的计算结果。随两相区配分时间延长,经IQ&P处理后试验钢的抗拉强度先增加后减小,而伸长率持续减小;在600 s时达到较好的强塑性匹配,强塑积为16 963.24 MPa·%。     

基于机器视觉的运煤车车号识别技术研究

在铁路运煤装车过程中为了快速、准确地识别车号,提出一种基于机器视觉的运煤车车号识别技术。将连通区域提取与投影分割法结合,实现车号的粗定位、细分割,并对图像中的断裂字符进行二次分割,构建了基于BP神经网络的分类模型进行车号识别,提升了煤炭装车的效率和精度。     

R2D2: A scalable deep learning toolkit for medical imaging segmentation

Deep learning has gained a significant popularity in recent years thanks to its tremendous success across a wide range of relevant fields of applications, including medical image analysis domain in particular. Although convolutional neural networks (CNNs) based medical applications have been providing powerful solutions and revolutionizing medicine, efficiently training of CNNs models is a tedious and challenging task. It is a computationally intensive process taking long time and rare system resources, which represents a significant hindrance to scientific research progress. In order to address this challenge, we propose in this article, R2D2, a scalable intuitive deep learning toolkit for medical imaging semantic segmentation. To the best of our knowledge, the present work is the first that aims to tackle this issue by offering a novel distributed versions of two well-known and widely used CNN segmentation architectures [ie, fully convolutional network (FCN) and U-Net]. We introduce the design and the core building blocks of R2D2. We further present and analyze its experimental evaluation results on two different concrete medical imaging segmentation use cases. R2D2 achieves up to 17.5× and 10.4× speedup than single-node based training of U-Net and FCN, respectively, with a negligible, though still unexpected segmentation accuracy loss. R2D2 offers not only an empirical evidence and investigates in-depth the latest published works but also it facilitates and significantly reduces the effort required by researchers to quickly prototype and easily discover cutting-edge CNN configurations and architectures.