Research on Segmentation of UAV Remote Sensing Imagery and Ground Objects Extraction
|School||PLA Information Engineering University|
|Course||Photogrammetry and Remote Sensing|
|Keywords||UAV Remote sensing Image Segmentation Ground targets extraction Object-oriented|
Low-altitude remote sensing system of unmanned aerial vehicles (UAV) , compared with the traditional aerospace remote sensing systems with flexible, fast and efficient , low operating cost , widely used , and many other significant advantages , especially in a small area of large scale diagram aspect has a very broad application prospects . Unmanned low-altitude remote sensing systems for high-resolution remote sensing images is an important expansion of traditional remote sensing image data source . Extracted from remote sensing images of ground targets , and has always been a challenging subject , UAV low altitude remote sensing images particularly rich in details of ground targets , focusing on the extraction method of image segmentation and object-oriented ground targets research . The main contents of this paper are as follows : 1) existing image segmentation algorithm , evaluation criteria and methods of image segmentation , to select the optimal edge detection operator , watershed algorithm , Otsu multi- threshold segmentation method and region growing four typical image segmentation algorithm , respectively UAV remote sensing images used in three different surface features cover types , using an empirical differences in evaluation methods to make a quantitative evaluation of UAV remote sensing image segmentation results , this evaluation method will reference results to compare the results obtained by the segmentation algorithm and the human eye split in order to achieve the performance evaluation of various segmentation algorithm ; 2) to explore the object-oriented image processing method using the decision tree method of object-oriented UAV remote sensing scale parameter extraction of images of ground targets and texture characteristics of the study results show that the decision tree method with object-oriented image processing method combination , can effectively simplify image processing workload , and improve the efficiency of extraction method of object-oriented ground targets ; 3 ) proposed an object-oriented UAV remote sensing image vehicle extraction method , decision tree training data extracted by the vehicle , UAV Remote Sensing Image in the extraction and classification of vehicles using the decision tree and auxiliary criterion , and on this basis , try to use continuous shooting UAV remote sensing image sequence to get the vehicle movement information .