期刊文献+
共找到3篇文章
< 1 >
每页显示 20 50 100
改进的DDeepLabV3+语义分割网络
1
作者 蔡思静 汪严昱 《福建理工大学学报》 CAS 2024年第1期95-102,共8页
针对语义分割网络在移动智能化终端上存在参数量大、分割精度不足的问题,提出一种改进的DDeepLabV3+网络算法。首先,采用深度可分离的MobileNet结构作为网络的骨干部分,降低网络的参数量和复杂度,从而有效减少了运行时间。其次,引入网... 针对语义分割网络在移动智能化终端上存在参数量大、分割精度不足的问题,提出一种改进的DDeepLabV3+网络算法。首先,采用深度可分离的MobileNet结构作为网络的骨干部分,降低网络的参数量和复杂度,从而有效减少了运行时间。其次,引入网络的低级特征,实现多尺度信息融合,减少网络下采样引起的空间信息损失。最后,结合注意力机制设计网络ASPP结构,增强特征提取在实验中的利用。优化后的网络结构在保持较高分类准确性的前提下,计算时间显著减少。网络的平均交并比在Cityscapes和Camvid数据集中分别提升了2.37%和2.13%。 展开更多
关键词 语义分割 SE注意力机制模块 DeepLabV3+网络
下载PDF
基于MEEMD与MP的电网低频振荡模态识别 被引量:10
2
作者 张程 刘佳静 +2 位作者 蔡思静 林谷青 匡宇 《智慧电力》 北大核心 2022年第4期89-95,共7页
针对目前电力系统低频振荡模态辨识的精确性和抗干扰性问题,提出了一种基于改进集合经验模态分解方法与矩阵束的电网低频振荡模态特征辨识新方法。首先利用改进集合经验模态分解方法将采集到的量测信号分解,从而获得若干个IMF分量序列... 针对目前电力系统低频振荡模态辨识的精确性和抗干扰性问题,提出了一种基于改进集合经验模态分解方法与矩阵束的电网低频振荡模态特征辨识新方法。首先利用改进集合经验模态分解方法将采集到的量测信号分解,从而获得若干个IMF分量序列及其残余量,再将剩余项去除后把其余本征模态函数进行重构,最后把重构信号通过矩阵束的分析来获知各个振荡模式信息。数值信号和EPRI-36节点系统的仿真验证了所提方法的有效性。 展开更多
关键词 低频振荡 MEEMD 矩阵束 排列熵 模态识别
下载PDF
3-D discontinuum numerical modeling of subsidence incorporating ore extraction and backfilling operations in an underground iron mine in China 被引量:1
3
作者 Huang Gang Kulatilake Pinnaduwa H.S.W. +2 位作者 Shreedharan Srisharan cai sijing Song Hongqing 《International Journal of Mining Science and Technology》 SCIE EI CSCD 2017年第2期191-201,共11页
An underground iron mine in China has been used as a case-study to research the subsidence due to ore extraction and backfilling during open stoping operations. A 3-D discontinuum numerical model was built incorporati... An underground iron mine in China has been used as a case-study to research the subsidence due to ore extraction and backfilling during open stoping operations. A 3-D discontinuum numerical model was built incorporating geologic complexities including faults and interfaces between different lithologies,and the stoping and backfilling sequence adopted from the mine plans. The stoping was carried out in two vertically stacked horizontal layers, with a total of 16 stopes. Large displacements of up to 50 cm were observed along the roof of the stopes, and a maximum surface subsidence of 22.5 cm was observed.Backfilling was found to eliminate subsequent displacements and subsidence. The extraction of the upper orebody was found to influence displacements in the lower orebody. Finally, a subsidence profile was constructed to show the subsidence at all locations along the length of the surface and region of influence on the surface. 展开更多
关键词 Underground iron mine Discontinuum numerical modeling Open stoping Backfill Subsidence
下载PDF
上一页 1 下一页 到第
使用帮助 返回顶部