Research On Aircraft Recognition Algorithm In Satellite Remote Sensing Images

With the development of remote sensing technology,the information in satellite remote sensing image is more and more abundant,the resolution is higher and higher,and the quality of remote sensing image is better and better.Such as cars,buildings,aircraft and other targets are clearly visible,which will facilitate the analysis of remote sensing images,and at the same time put forward higher technical requirements for remote sensing image analysis.the segmentation,detection and identification of aircraft targets in satellite remote sensing images are of great significance in both military and civilian fields.This paper mainly focuses on the segmentation,merging and recognition of aircraft targets in satellite remote sensing images.The main research work is as follows:(1)Based on the experimental realization of existing algorithm(the between-cluster variance method,artificial segmentation threshold segmentation method and K-means segmentation method)was carried out on the satellite remote sensing image plane target segmentation,and on the analysis of the advantages and disadvantages of the existing algorithm is put forward after an automatic threshold segmentation algorithm,in terms of aircraft targets,such as the body material,coating,radian caused the fuselage to high light reflectivity,present the bright white,on this basis,through the statistics of gray level from 0to 255 every grayscale appear frequency,first find out the grayscale frequency and the corresponding maximum grey value as a wave,Then,after detecting the first peak after the maximum peak,the gray value corresponding to the valley between the two peaks is taken as the image segmentation threshold,and a better segmentation effect is obtained.(2)Using eight connected domain mark depth first search algorithm of automatic threshold segmentation to target combined binary image,in the binary image after segmentation first foreground pixels as the seed point and then constantly to seed point spread around,all the pixels to meet the conditions,until there is no pixels to satisfy the conditions and the stack is empty,finished the first sign of connected domain at this time.When the next foreground pixel is found,the method and condition are consistent with the first connected domain marker method,and so on,until the entire image is traversed and all connected domain markers in the image are completed.(3)Through the automatic threshold segmentation and mark the connected domain algorithm to target after the merger,the SIFT algorithm,Fourier shape descriptor,HU invariant moments after the merger deal with target and the target repository prior target matching recognition,all have to recognize when an object’s ability to translation,rotation and scaling occurs,through comparing the recognition algorithm of recognition rate experiment,SIFT algorithm,image matching,most powerful recognition rate reached 100%,Fourier descriptor,HU moment invariants recognition,the recognition rate is 93%.