联合像素级和对象级分析的遥感影像变化检测

发布时间：2018-06-23 11:01

本文选题：像素级 + 对象级　；参考：《测绘学报》2017年09期

【摘要】：为改善高空间分辨率遥感影像的变化检测精度,提出一种联合像素级和对象级分析的变化检测新框架。首先将多时相影像进行叠合,对叠加影像进行主成分分析,并利用基于熵率的方法对第一主成分影像进行分割,通过改变超像素数目来获取多层次不同尺寸大小的超像素区域。同时,对多时相影像进行光谱差异和纹理差异分析,采用自适应PCNN神经网络方法进行图像融合,利用水平集(CV)方法对融合后的影像进行分割获取像素级变化检测结果。最后,结合多尺度区域标记矩阵对检测结果进行变化强度等级量化和决策级融合,作为变化检测的后处理部分,以获取最终的对象级变化检测结果。采用SPOT-5多光谱影像进行试验。结果表明这种新框架可以有效集成基于像素和基于对象两种图像分析方法的优势,能够进一步提高变化检测过程的稳定性和适用性。
[Abstract]:In order to improve the accuracy of change detection for high spatial resolution remote sensing images, a new change detection framework combining pixel level and object level analysis is proposed. Firstly, the multi-temporal image is overlapped and the superposition image is analyzed by principal component analysis (PCA). The first principal component image is segmented based on entropy rate, and the multi-level and different size super-pixel regions are obtained by changing the number of super-pixels. At the same time, spectral and texture differences of multitemporal images are analyzed, adaptive PCNN neural network method is used to fuse images, and level set (CV) method is used to segment the fused images to obtain pixel level change detection results. Finally, the multi-scale regional marker matrix is used to quantify the change intensity level and the decision level fusion, which can be used as the post-processing part of change detection to obtain the final object-level change detection results. SPOT-5 multispectral image was used to test. The results show that the new framework can effectively integrate the advantages of two image analysis methods based on pixel and object, and can further improve the stability and applicability of the change detection process.
【作者单位】：武汉大学测绘遥感信息工程国家重点实验室;
【基金】：国家重点研发计划(2016YFB0502600) 测绘遥感信息工程国家重点实验室开放基金(16E01) 国家自然科学基金(41471354)~~
【分类号】：TP751

【相似文献】