8334

7732

كارشناسي ارشد

مخابرات

اصفهان: دانشگاه صنعتي اصفهان، دانشكده برق و كامپيوتر

1392

نه،92ص.: مصور،جدول،نمودار

ص.ع.به فارسي و انگليسي

بهزاد نظري، رسول امير فتاحي

3/10/92

مهندسي برق و كامپيوتر

ID7732

به فارسي و انگليسي: قابل رويت در نسخه ديجيتالي

Target Identification in Aerial Images by DOOS New Descriptors Gholamreza Ramezanian g ramezanian@ec iut ac ir Date of Submission 2013 05 11 Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Degree M Sc Language FarsiSupervisor Behzad Nazari Rasul Amirfattahi nazari@cc iut ac ir fattahi@cc iut ac irAbstract UAV is now widely used for aerial image acquisition transmission and recognition One of the UAVimages research object is target identification In this thesis proposed an algorithm for target identificationbased on new feature descriptors The new descriptors are result of the Dominate Orientation of Subregions so called DOOS In this algorithm after detecting target and scene image keypoints and describing them byDOOS corresponding keypoints between to image is discovered By RANSAC algorithm removed the outlierand affine transform is estimated At the end by affine transform target location is found in target image Theexperiments show that DOOS descriptors are more robust than SIFT and BRISK In additional because ofthe low dimensionality and structure of DOOS descriptors matching of them is faster than other descriptorslike SIFT and SURF We propose a simple implementation for matching DOOS descriptor to show it is easyto implement on DSPs and FPGAs Keywords Target Identification DOOS descriptor SIFT Aerial images

بهزاد نظري، رسول امير فتاحي