融合多特征的对象级航空遥感图像变化检测

Object-oriented change detection by combining multi-features for aerial remote sensing imagery

针对航空遥感图像,构建一种面向对象的融合JS(Jensen-Shannon)散度特征与互相关特征的变化检测算法。首先,应用多尺度分割算法获取像斑;然后,提取反映像斑内像素灰度分布的总体统计特征的JS散度以及反映像斑内部结构的变化特征的互相关特征,应用决策级融合方案对两个优势互补的特征进行有效融合,进而探测变化区域;最后与固定权重融合的检测结果进行精度对比。结果表明:本文方法的平均检测精度达到93.07%,误检率平均为7.13%,漏检率平均为4.37%,比仅基于JS散度特征、互相关特征、固定权重融合的检测方法精度分别提高了8.98%、4.71%和4.20%。因此,该变化检测方法不仅能有效提取变化区域,而且提高了变化检测的精度,在航空遥感图像变化检测中具有有效性与应用潜力。

An object-oriented change detection algorithm was proposed by combining JS(Jensen-Shannon)divergence feature and cross-correlation feature for aerial remote sensing imagery.First,multiscale segmentation was employed to get image objects.Second,JS divergence,which reflects the overall statistical features of gray distribution in each object,and cross-correlation feature,which describes the internal structure changes of each object,was extracted.Third,the decision-level fusion algorithm was applied to fuse the two complementary features to detect the changed area.Finally,the accuracy of the results was compared with that of the fixed-weight fusion algorithm.Experimental results indicate that the average accuracy of the proposed method reaches 93.07%,the average false detection rate and omission rate are 7.13%and 4.37%and that the average accuracy of the proposed method is 8.98%,4.71%and 4.20%higher than those of the algorithms based on JS divergence feature,cross-correlation feature and fixed-weight fusion,respectively.Hence,the proposed method can not only detect the change area effectively,but also improve the accuracy of change detection,which shows the potential and effectiveness of the proposed method in change detection of aerial remote sensing images.