基于神经网络的遥感影像分类方法研究

本文在研究BP神经网络和模糊理论的基础上,提出了传统BP算法的一种改进方法和基于模糊系统的神经网络遥感影像分类方法。通过试验表明:基于模糊技术的神经网络分类方法要优于BP神经网络方法,取得了令人满意的效果。     

一种深度图像拟合圆柱的优化方法

圆柱状构件广泛存在于古建筑仿真中,为了提高古建筑深度图像圆柱拟合的精度,针对古建筑仿真3维建模的特点,提出基于高斯图获取圆柱拟合参数初始值,然后利用圆柱距离函数参数化法进行非线性最小二乘拟合的圆柱拟合方法。实验证明,该方法不仅具有较高的精度,而且拟合结果稳定可靠。     

《遥感图像处理》课程教学在地理学专业中的探索与实践

针对地理专业的自身特点和当前遥感图像处理的教学现状，提出了遥感图像处理课程在地理专业中的主要教学内容框架、教学方法以及教学改革方向，并对该课程的实践教学部分做了重点介绍，最后论述了该课程在地理学专业实践教学中的取得一些教改成果。     

多源遥感影像配准技术分析

图像配准技术是近年来发展迅速的图像处理技术之一,是图像融合、目标提取与定位、变化监测、高分辨率影像重建等处理工作不可缺少的步骤。首先给出了多源遥感影像配准的概念和基本过程,然后着重从匹配基元、转换函数、相似性测度、匹配策略4个方面对国内外现有的影像配准技术和方法进行了分析和评述,最后指出了影像配准技术所面临的主要难题和发展方向。     

Gibbs sampler by sampling-importance-resampling

Among the Markov chain Monte Carlo methods, the Gibbs sampler has the advantage that it samples from the conditional distributions for each unknown parameter, thus decomposing the sample space. In the case the conditional distributions are not tractable, the Gibbs sampler by means of sampling-importance-resampling is presented here. It uses the prior density function of a Bayesian analysis as the importance sampling distribution. This leads to a fast convergence of the Gibbs sampler as demonstrated by the smoothing with preserving the edges of 3D images of emission tomography.     

IKONOS影像的定向精度研究

简要介绍了IKONOS卫星遥感影像适用于生产地理信息产品的特点，通过生产试验研究详细分析了IKONOS影像作为数据源在生产地理信息产品时的定向精度问题。     

高光谱影像数据预处理技术评价

高光谱影像数据预处理是构建高效的影像数据库以及实现地理目标成分信息反演与地理目标识别的基础。本文从反射率演算、光谱微分技术以及高光谱影像数据的降维处理等三个方面对高光谱影像数据预处理进行了研究。实现了基于高光谱影像的反射率反演算法，针对高光谱遥感数据影像分辨率比较低的现状，提出了直方图均衡化、色调平衡、对比度调整等影像处理算法，并给出了实验结果。     

基于小面元相对配准的正射影像快速更新

本文提出一种基于小面元相对配准的正射影像快速更新方法，该方法不涉及常规摄影测量的处理方法，而是直接利用新获取的卫星遥感影像和原始正射影像进行高精度小面元相对配准，然后将卫星遥感影像纠正到原始正射影像参考坐标系，进而实现正射影像的快速更新。     

On the flow structure and turbulence during sweep and ejection events in a wind-tunnel model canopy

Particle image velocimetry (PIV) data obtained in a wind-tunnel model of a canopy boundary layer is used to examine the characteristics of mean flow and turbulence. The vector spacing varies between 1.7 and 2.5 times the Kolmogorov scales. Conditional sampling based on quadrants, i.e. based on the signs of velocity fluctuations, reveals fundamental differences in flow structure, especially between sweep and ejection events, which dominate the flow. During sweeps, the downward flow generates a narrow, highly turbulent, shear layer containing multiple small-scale vortices just below canopy height. During ejections, the upward flow expands this shear layer and the associated small-scale flow structures to a broad region located above the canopy. Consequently, during sweeps the turbulent kinetic energy (TKE), Reynolds stresses, as well as production and dissipation rates, have distinct narrow peaks just below canopy height, whereas during ejections these variables have broad maxima well above the canopy. Three methods to estimate the dissipation rate are compared, including spectral fits, measured subgrid-scale (SGS) energy fluxes at different scales, and direct measurements of slightly underresolved instantaneous velocity gradients. The SGS energy flux is 40–60% of the gradient-based (direct) estimates for filter sizes inside the inertial range, while decreasing with scale, as expected, within the dissipation range. The spectral fits are within 5–30% of the direct estimates. The spectral fits exceed the direct estimates near canopy height, but are lower well above and below canopy height. The dissipation rate below canopy height increases with velocity magnitude, i.e. it has the highest values during sweep and quadrant 1 events, and is significantly lower during ejection and quadrant 3 events. Well above the canopy, ejections are the most dissipative. Turbulent transport during sweep events acts as a source below the narrow shear layer within the canopy and as a sink above it. Transport during ejection events is a source only well above the canopy. The residual term in the TKE transport equation, representing mostly the effect of pressure–velocity correlations, is substantial only within the canopy, and is dominated by sweeps.