基于特征增强与最小误差分割的变化检测方法

Change detection method based on feature enhancement and minimum error threshold segmentation

以河北省石家庄市2003年和2004年的专题制图仪(TM)遥感影像为例,针对各波段光谱特征,提出了一种基于地物特征增强的变化检测方法。在两期影像上对各类地物采样并计算样本在不同波段的均值、标准差等特征量,以确定波段组合运算的加权系数,计算特征增强图像,实现两期影像中所指定地物类型的特征增强;计算两期特征增强影像的差异影像;使用最小误差分割法获取变化检测结果。通过对比实验可知:方法提取变化区域总体精度达到90%,相对于传统的基于主成分分析(PCA)的变化检测方法,具有较高的检测精度,较好的可行性与适应性。

Take thematic mapper（ TM） remote sensing image of Shijiazhuang,Hebei Province in 2003 and 2004 as example,aiming at spectrum characteristics at each waveband,a change detection algorithm is proposed,which is based on surface feature enhancement technology. Firstly,in order to determine the weight coefficient of band combination operation,the average,standard deviation and other characteristic values of different bands are calculated by ground samples from multi-temporal images,and then the feature enhancement image can be calculated. Secondly,the difference image of two phases feature enhancement images is calculated. Finally,the minimum error threshold segmentation is performed on the difference image to obtain the change detection results.It shows through comparison experimental results that the overall precision of distinguishing reaches 90 %,and compare with the traditional change detection method based on principal component analysis（ PCA）,the new method has higher precision,better properties of applicability and feasibility.