Buildings Change Detection Based On Shape Matching For Multi-resolution Remote Sensing Imagery

Buildings change detection has the ability to quantify temporal effects,on urban areas,for urban evolution study or damage assessment in disaster cases.In this context,changes analysis might involve the utilization of the available high resolution satellite images with different scales for quick responses where decision makers need fast support tool to provide a precise monitoring of area of interest.However,before starting change detection process using traditional pixel based methods,image resampling must be applied on the images in order to produce the same pixel size and allow comparing multi-date images to establish change detection map.Therefore,image resampling might affect the spectral information of pixels and small details present in the original image might be lost in the resampled image which affects the accuracy of change detection results.Also,traditional change detection methods based on pixels are inadequate due to the low separability between image classes.Hence,to avoid using traditional methods with image resampling outcomes and salt-pepper effect,buildings change detection based on shape matching is proposed for multi-resolution remote sensing images.Because the object's shape can be described and extracted from remote sensing imagery and corresponding objects in multi-resolution images have the same geometry and position,it is useful to detect buildings changes using shape analysis.Therefore,the proposed approach can identify new and demolished buildings with their class transition in urban areas using geometric properties of objects of interest as follow:(1)After rectifying the desired multi-dates and multi-resolutions images,by image to image registration with optimal RMS value,objects based image classification is performed to extract objects shape from the images including buildings.(2)Next,Centroid-Coincident Matching is performed on the extracted building polygons,based on the Euclidean distance measurement between shapes centroid(from shape To to shape Ti and vice versa),in order to define corresponding objects' shapes with minimum distances.Then,New and Demolished buildings are identified based on the obtained minimum distances those are greater than RMS value(No match in the same location).(3)Finally,"From-To" changes can be implemented by defining the closest polygon,of each extracted class objects shape,to the obtained changed buildings polygon and performing polygon intersection(overlapping)to describe class transition over time.(4)In order to verify the effectiveness of this method,the accuracy assessment of the detected changes was carried out based on the accuracy measures using ground truth changes map which was generated by manually on screen digitization.Also,the results of buildings change detection by post-classification comparison of the images,after image resampling,are used to compare the accuracy of the obtained change detection results.The obtained experimental results demonstrate the effectiveness of the proposed change detection approach to detect New and Demolished buildings with an overall accuracy of 97%and a kappa value reach 0.94.These results show that the presented method in this study can be efficiently used to create accurate buildings change detection maps and update geodatabases for urban environment monitoring.