Font Size: a A A

Research On Target Recognition For Autonomous Landing Of Ship-board Rotor Unmanned Aerial Vehicles

Posted on:2014-09-10Degree:MasterType:Thesis
Country:ChinaCandidate:Y Y ShiFull Text:PDF
GTID:2268330422952860Subject:Navigation, guidance and control
Abstract/Summary:PDF Full Text Request
Rotorcraft UAV has many advantages like small size, low cost and easy to carry, so it has a broadapplication prospects in many areas.This paper has a research on target recognition for autonomouslanding of ship-board Rotor Unmanned Aerial Vehicles.We set a artificial object on the runway of theship,then the object can be recognized by using computer vision, ultimately UAV can preciselynavigate and land automatically.First, a threshold segmentation method based on the HSI space is chosen by making full use ofcolor information of the object. The segmentation threshold is determined through statisticalhistogram by collection of sample in plurality sets of light condition, considering for uneven lightingconditions at the sea surface.Then morphological filtering is used on the divided image,and finally,edge detection and contour tracking method are used to get the contour chain code of the objectpreparing for the next feature extraction.Secondly, affine invariant moment is chosen as the invariants features according to the specificlanding scene requirements,and take the real-time for consideration, the invariant moments calculationmethod based on contour chain is presented in this paper,then simulation experiments are designed forcomparation of methods in this paper with traditional methods from two aspects,that is computationtime and accuracy.The results show that this algorithm is real-time and had a high accuracy.Then, recognition classifier is designed using Support Vector Machine, and multiple samplesunder natural conditions are collected as sample libraries, then in order to verify the effectiveness ofthe proposed algorithm,some offline recognition experiments are built.The results show that thisidentification scheme can effectively identify the feature identification, and can meet the accuracyrequirements.Finally, position and attitude estimation is preliminarily studied after identify the characteristics,and the point estimates of the object is given, finally the calculation method of relative position andattitude parameter required in the landings is drawed.
Keywords/Search Tags:Rotorcraft UAV, Autonomous landing on the ship, Affine moment invariants, SupportVector Machine, Target Recognition
PDF Full Text Request
Related items