结合空间上下文与慢特征分析的多光谱影像变化检测

为了提高多光谱影像变化检测的精度,本文提出了一种结合空间上下文与慢特征分析的方法。首先采用自适应空间上下文提取算法围绕像素构建自适应区域,探索像素周围的上下文信息;然后通过迭代慢特征分析,由相应像素周围的成对自适应区域定量计算成对像素之间的变化强度,增强变化区域与未变区域的可分性;最后生成变化强度图像,采用大津阈值法作二值分类,将变化强度图划分为二值变化检测图。利用Landsat 7卫星ETM+传感器的图像,与4种基于代数的方法及基于变换的方法进行对比试验,结果表明,本文方法在降低漏检方面有所改善,提高了召回率。

Multispectral image change detection based on spatial context and slow feature analysis

To improve the accuracy of multispectral change detection, a method combining spatial context and slow feature analysis is proposed. First, an adaptive spatial context extraction algorithm is used to construct an adaptive region around the pixel to explore the context information around the pixel. Then through the iterative slow feature analysis, the change intensity between paired pixels is quantitatively calculated from the paired adaptive region around the corresponding pixel. The separability of the changed area and the unchanged area is enhanced. Finally, the change intensity image is generated, and the Otsu threshold method is used for binary classification, and the change intensity map is divided into binary change detection maps. Experiments use images from the Landsat 7 satellite TM sensor to compare with four algebra-based and transformation-based methods. The results demonstrate that the method in this paper performs better in terms of reducing omission errors and improving recall rate.