遥感影像地物特征识别的频谱能量分析方法探讨

从Parseval能量守恒定理和地物特征的图谱识别理论出发,对遥感影像地物特征识别的频谱能量分析方法进行了分析。通过分析不同频谱半径范围内所占的能量比重,初步考察了地物特征频谱能量的分布特点。进而通过对频谱能量环状和楔状采样,定量描述了表征地物粗糙度、周期性及方向性特征的频谱能量分布状况。最后以居民楼地物为例,根据建立的高、低频识别标志,结合在频谱能量上具有方向和频带选择性的匹配Gabor滤波器,实现了居民楼地物对象的提取。实验结果表明,此方法可以获取在空间域中无法得到的地物信息表征,为进一步开展地物特征的识别和提取提供了一条新思路。

Applying Frequency Spectrum Energy Analysis Theory and Methodto Recognize Objects for Remote Sensing Image

Relying on the Parseval energy conservation theorem and‘Theory of Tupu’，thefeasibility of object recognition methods based on‘pixel-baesd spectral energy’to‘feature-based spectrum energy’was demonstrated.A preliminary investigation of the percentage ofspectral energy contained by circles of different radius taking total energy,and the spectralenergy distribution was undertaken.Meanwhile,aquantitative analysis of coarse-grainedlevel,cyclical,and directional information was studied using wedge-shaped and ring-shapedenergy spectrum sampling methods.Finally,based on low-frequency and high-frequency rec-ognition marks,residential buildings were taken as an a test-case to assess this theory andmethod.Building objects were extracted with matched Gabor filters having direction and fre-quency selectivity.The results indicate that this method provides a new and effective ideameans further identify and extract information from high resolution imagery.