高分辨率遥感影像变化检测的频域分析预处理方法

A Preprocess for Change Detection Method from High Resolution Remotely Sensed Imageryin Frequency Domain

目前,大多数高分辨率遥感影像变化检测方法的效率不高,精度也难以达到实际应用要求,本文引入频域分析的方法进行变化检测预处理,达到提高变化信息检测精度和效率的目标。首先,对影像频域能量的分布特征进行分析,并以此为基础,利用累加距离匹配函数,完成试验影像的合理分块;其次,通过变化性纹理规则度的定义和计算,实现对同能量下变化信息的精确描述,缩小变化区域搜索范围,有效抑制背景信息对目标信息的干扰。最后,通过3种常用的变化检测方法的验证,发现变化检测预处理方法可以使检测算法在现有精度和效率的基础上有进一步提高。

At present,the efficiency of change detection from high resolution remotely sensed imagery is generally not very high, and accuracy couldn＇t meet the practical requirements.A method which introduced frequency domain analysis is put forward in this paper.Firstly, it studies the distribution of image frequency energy.Using it as a basis, the image division is accomplished by summed-up distance matching function.Secondly, the definition and computation of texture regularity of change information（TRCI） clearly and accurately expresses the change information according to equal energy rule of image blocks, greatly narrowing the search area.Thus, the background information could be greatly suppressed so as to improve the efficiency and precision of detection.Finally, through measuring and comparing of the accuracy and efficiency of three different detection algorithms before and after introducing the preprocess, we find that it could provides more accurate results on the existed basis.