Font Size: a A A

Research On Detection And Track Methods Of Flexible Robotic Fish

Posted on:2018-01-30Degree:MasterType:Thesis
Country:ChinaCandidate:C GuoFull Text:PDF
GTID:2348330533469948Subject:Mechanical engineering
Abstract/Summary:PDF Full Text Request
With the development of robot exploration,it becomes a fashion to learn f rom animals.After experiencing one billion years evolution,fish has gained sup erior swimming performance,and becomes the interest of many researchers.So it is a essential need to pursue the features of fish in the filed of intelligent m achine.But in the market,many robotic fishes can not swim like real fish sinc e researchers have not master the laws of fish efficient swimming.It is helpful to study the relationship between the swimming speed of fish and the frequenc y of tail beat.With the increasingly development of computer vision technology,it becomes an efficient way to measure the swimming speed of fish.Many res earchers are intended to use this way.However the motion of many creatures i s uncertain when they swim,so it becomes a very difficult problem for many s cientist.In order to measure the motion of fish,it is important to differ the fi sh from background and capture its profile.Nowadays the measure method basi ng two eyes vision becomes an efficient way in the field of the motion of fish.So it becomes a focus in the filed of measurement and acquisition.This paper firstly creates a platform of embedded measurement and transpla nting the library of OPECV and QT.Then the theory of two camera and the e stablishment of vision system is explored.Then the detailed explanation of the detection of the moving objects and capture of moving objects are given.The c amera calibration,picture correction,the separation between the object and back ground,and a series of picture pre-treatment technique are accomplished.Secondly the algorithm of the detection and capture of the moving objects are studied in detail.In order to choose an appropriate algorithm,many algorith ms are compared and integrated together.The Fast-RCNN,Kalman filter trackin g algorithm,MeanShift tracking algorithm are chosen.Using them can obtain th e relationship between the swimming speed and the beat frequency of a robotic fish.Finally,the algorithm developed in PC are transplanted into the embedded system.In the platform of arm,the algorithm is improved.It can improve the demand of real-time.The algorithm is verified.Its feasibility and accuracy can be verified through experiments.
Keywords/Search Tags:Robotic fish, stereo vision, Kalman filtertracking algorithm, ARM
PDF Full Text Request
Related items