Automated Extraction of Urban Water Bodies from ZY‐3 Multi‐Spectral Imagery

摘要: The extraction of urban water bodies from high‐resolution remote sensing images, which has been a hotspot in researches, drawn lot attention both domestic and abroad. A challenging issue is to distinguish the shadow high‐rise buildings bodies. To tackle this issue, we propose automatic method (AUWEM) extract images. First, order improve accuracy, refine NDWI algorithm. Instead Band2 NDWI, select first principal component after PCA transformation as well Band1 for ZY‐3 multi‐spectral image data construct two new indices, namely NNDWI1, sensitive turbid water, NNDWI2, body whose spectral information interfered by vegetation. We superimpose threshold segmentation results generated applying NNDWI1 then detect remove shadows small areas using object‐oriented detection technology, finally obtain extraction. By comparing Maximum Likelihood Method (MaxLike) find that average Kappa coefficients AUWEM, MaxLike five experimental are about 93%, 86.2% 88.6%, respectively. AUWEM exhibits lower omission error rates commission compared with MaxLike. total three methods 11.9%, 18.2%, 22.1%, not only shows higher edge but it also relatively stable change threshold. Therefore, can satisfy demands extracting