Research On Change Detection Method Based On Spectral Trend Feature And Spatial Information

Change detection based on multi-temporal remote sensing image has become an important means to obtain land cover change information quickly and widely.It plays an important role in urban planning,smart agriculture,forest land monitoring,water area management,disaster assessment,natural resources census and other fields.Because optical remote sensing images have the characteristics of good visualization,periodic collection and abundant information,change detection based on optical remote sensing images has been a hot topic in research.The improved spatial resolution and limited spectral resolution leads to the common phenomenon of "the same objects with different spectrum" and "the different objects with same spectrum",and it is difficult to obtain ideal detection results based on the single spectral feature.To solve the above problems,this thesis studies the change detection method based on the combination of spectral trend feature and spatial information,in order to improve the accuracy and reliability of change detection.The specific contents are as following:(1)For pixel-level methods,the difference information is often extracted directly from gray values,ignoring the correlation of spatial relationship.In order to solve the above problem,this thesis proposes a change detection method based on spectral trend information.By introducing the concept of spectral trend and integrating spectral trend feature and spatial neighborhood information,the reliability of spatial information is enhanced.Experimental results demonstrate that the proposed method can reduce false detection error effectively and achieve higher detection accuracy,but also it is robust to noise.(2)Aiming at the phenomenon that the detection effect of object-level change detection method is greatly influenced by the segmentation scale factor,a change detection method based on the combination of statistical region merging algorithm and region detection operator is proposed.The regional object is constructed according to pixels,which fully reflects the features of ground objects and makes up for the errors caused by the uncertainty of segmentation factors in object extraction and change detection.Experimental results show that by introducing the region detection operator,the accuracy of spatial information is effectively improved,and the accuracy of change detection is also improved.There are 32 figures,15 tables and 94 references in this thesis.