排序方式： 共有3条查询结果，搜索用时 62 毫秒
1
1.
In the present study, three wavelet basis functions (Mexican-hat, Morlet, and Wave) were used to analyze atmospheric turbulence data obtained from an eddy covariance system in order to determine effect of six meteorological elements (three-dimensional wind speed, temperature, and CO2 and H2O concentrations) on the time scale of coherent structures. First, we used the degree of correlation between original and reconstructed waveforms to test the three wavelets’ performance when determining the time scale of coherent structures. The Wave wavelet’s reconstructed coherent structure signal best matched the original signal; thus, it was used for further analysis of the time scale, number, and time cover of the meteorological elements. We found similar results for all elements, though there was some internal variation, suggesting that coherent structures are not inherently dependent on these elements. Our results provide a basis for proper coherent structure detection in atmospheric turbulence and improve the understanding of similarities and differences between coherent structure characteristics of different meteorological elements, which is helpful for further research into atmospheric turbulence and boundary layers. 相似文献
2.
基于图像特征的地类识别技术精度低且复杂度高,难以满足土地利用动态监测的实时性要求。而基于深度学习的遥感影像地类识别技术数据处理及特征提取能力较强,能够有效提升识别精度,使地类信息获取更加智能化,因而被广泛应用于遥感影像地类处理。根据地类识别技术不同,可以分为遥感影像分割、遥感影像分类以及目标监测三种方式。每类识别技术根据训练数据是否有标注可以分为有监督学习、半监督学习以及无监督学习三种类型。通过对相关技术和文献的调研和分析,介绍各种深度学习模型的理论基础与基本结构,总结国内外基于深度学习的遥感影像地类识别方法的优缺点,并指出未来的发展方向。 相似文献
3.
遥感影像变化图斑智能化提取是自然资源动态监测工作的基础。本文简述了遥感影像变化检测技术演进历程及特点,提出同时使用ResNet、U-Net和孪生神经网络3种深度学习算法,设计了集“影像预处理、智能提取、协同筛查”于一体的遥感影像变化图斑智能化提取平台,并详细阐述了各功能模块设计思路。实践表明,融合3种深度学习算法有利于解决单一深度学习网络模型改造难度较大、适用范围有限等难题,有效提升了遥感影像变化检测的查全率,工作效率比目视解译提升超过3倍。研究成果已在湖南省自然资源“1+N”卫星监测工作中广泛应用。 相似文献
1