Sub－pixel 　Location　of 　Feature　Point　Based　on Taylor 　Expansion　and　Its 　Application

Considering　the　hi gh　demand　of 　interest 　pointspositioning　precision 　for 　aerotriangulation　of UAV　imagesthis　paper 　proposes 　an 　improved 　sub－pixel 　location　method　based　on　Harris　al gorithm． Firstl yHarris　al gorithm　was　used 　to　detect 　the　ori gin 　interest 　points．Secondl ya　least 　square 　fittin g of　Taylor 　expansion　to 　the 　image　gre y　scale 　surface　was　done　around 　the　nei ghborhoods　of 　each　ori gin interest 　point 　and　a　Gaussian－distance 　al gorithm　was　emplo yed 　to　determine 　the　weight．Thirdl ysubpixel 　location 　of 　the 　interest 　point　was　obtained　by　calculating　the 　extremities　of 　the　derived 　function．Tests　were　done 　for 　ei ght 　kinds 　of 　computer－simulated 　images 　and　three　kinds 　of 　actual　UAV　images experimental 　results 　show　that 　our　proposed　method　has 　the　desired　efficienc y　and　noise 　immunity the　positioning　accurac y　can　meet　the 　re quirements 　of 　aerotriangulation　of　UAV　images