Change Detection Algorithm Based on Discrete Wavelet Transforms and Neighborhood Fuzzy C-Means

基于离散小波变换和邻域模糊C均值的变化检测方法

Received: 12 April 2017

Abstract

A new algorithm on Discrete Wavelet Transform (DWT) and neighborhood FCM is proposed to detect change area from remote sensing image. First, the subtraction and ratio image are obtained by the subtraction and ratio method from the two registered remote sensing images; Then, the DWT is applied to the subtraction and ratio image, the region intensity-based and energy-based fusion rules is adopted to the low frequency and high frequency wavelet coefficients, and the inverse DWT is used to obtain the final difference image; At last, the neighborhood FCM is carried out to get the change areas, the spatial distance information and gray difference information are considered in the objective function of FCM, which could avoid misclassification and enhance the detection probability. Experimental results show that the proposed algorithm has strong ability to suppress noise and good detection results; the detection probability of unban change area can reach to 98.45%, whereas, the detection probability is up to 87.5% for the discontinuous forest change area.

摘要

提出了一种基于离散小波变换和邻域模糊C均值（FCM）的变化检测方法。首先，采用差值法和比值法获取已配准后的两幅遥感图像的差值图和比值图；其次，对求取的差值图和比值图进行离散小波变换，采用基于区域强度信息和区域能量信息的融合规则分别对低频带小波系数和高频带小波系数进行融合，并采用离散小波逆换获取最终的差异图像；最后，采用基于邻域FCM的方法从差异图像中检测出变化区域，提出了把空间距离信息和邻域灰度差值信息引入到FCM的目标函数中，以避免误分类、提高检测概率。实验表明，所提出的方法具有较强的抑制噪声的能力和较高的检测概率，对城市面积变化检测概率达到了98.45%，对于变化区域不连续的森林面积变化的检测概率也达到了87.5%。