The Stereoscopic Detection Of Building Changes From Optical Image And Airborne Laser Scanning

With the rapid development observation of remote sensing technology, it ispossible that we can repeate to observe the same urban area soon and more and morekinds of remote sensing data can be got such as the high-resolution remote sensingimages and airborne lidar data. Change detection technology obtains the changes ofbuilding accurately by processing and comparing the difference of multi-temporalremote sensing data in the same area. The building is the most important symbol ofthe city. It is significant to detect changes accurately in the building of urban areasand achieve the qualitative visualization and quantitative analysis of the changesultimately.Firstly, it’s difficult to achieve high accuracy for registration and radiometriccorrection between the multi-temporal high-resolution remote sensing images and anew change detection method is proposed based on the cross correlation coefficient.After obtaining the cross-correlation image, the two-dimensional Ostu algorithm isused to set the threshold automatically and get the change mask. In order to optimizethe change mask, we make the filling, opening and closing operations and acheivequantitative analysis of building changes.Secondly, the main change of building is that the height change of the building,the height of DSM data is independent of lighting condition and we can detect thechange of building by using the height difference of DSM data. The heightdifferencing of the multi-temporal DSM data is positive or negative that can dividethe change type of building into two parts, which are the dismantled building and thenew building. To solve the problem that the changes of building is easy to beinterfered by the non-construction such as trees, we propose a method based onregion and line features to optimize the change and remove the interference ofnon-construction effectively. The result of change detection is also consistent withthe measured Data.Finally, since that the fine reconstruction data can contain whole shape andstructure information of buildings and show the change in the sides of buildings, thealgorithm based on the minimum distance differencing adaptive threshold isproposed. The multi-temporal reconstruction data need to be registered to the same3D coordinate system to be convenient for the subsequent detection. Due to thedifference between regions of the reconstruction data, it is difficult to set thethreshold of distance. Two methods adjusting threshold are proposed, which arefixed threshold and local threshold.Achieve detection about two types of building change, obtain the change point cloud and extract the three-dimensional featuressuch as the volume, center, moments. It proves that the change detection based onreconstruction data can find the side change of building with a higher accuracy andachieve the change detection of buildings in the three-dimensional space. The3D features extraction based on3D point cloud data, especially the vluome arealmost Consistent with the measured data, which proves the accuracy of thealgorithm performance.In a word, this paper starts research with optical image，DSM data and finereconstruction data obtained from LIDAR point cloud and the information used todetect changes is the gray, the elevation and then three-dimensional coordinatedistance，which make the change detection of building in the three-dimensionalstructures more accurate.