Research And Realization Of The Multi-target Tracking System For Robotic Fish Based On Global Vision

In recent years, the research of underwater bio-mimetic robotic fish has become a hotfocus at home and abroad. Many countries and regions, such as the United States, Europe andJapan, have developed large-scale studies in the bio-mimetic robotic fish and carried out a lotof research results. Bio-mimetic robotic fish combines the technique from electronics,communication, automation, artificial intelligence, computers and other key technologies, atthe same time it provides a good experimental and testing platform for the theoretical researchof related areas，such as image processing, fuzzy control, group formation, turbulence.Visual subsystem is a very important part of multi-robotic fish collaboration controlsystem. The rapid detection and tracking of moving robotic fish is very significant whencontrolling the robotic fish based on global vision. The control system collects the real-timepanoramic image through the visual subsystem. After a series of image processing, the controlplatform can get the positional information of the targets, which is the only informationalsource when the host computer is analyzing and deciding and controlling the bio-mimeticrobotic fishes’swimming.After researching for some problems about the realization of the visual subsystem, thispaper presents a target tracking algorithm which is based on matching template, sums up andputs forward some significant ways and techniques of the hardware and software in therealization of visual subsystem.The main research results in this paper are as follows:Firstly, this paper makes a detail description on the visual subsystem’s hardware andsoftware of the multi-robotic fish collaboration control system and presents the solution ofsome problems in the system, such as image distortion.Secondly, a new scheme is presented in the target poses acquirement process of thevisual subsystem, in which a video tracking method is used to replace the former targettracking with the method of color matching steps. And then, with the research and comparisonof the current template matching algorithms and considering some characteristics ofbio-mimetic robotic fish, this paper uses the dual-template matching algorithm for targettracking to get targets’ pose information.Thirdly, considering the shape characteristics of bio-mimetic robotic fish itself anddepending on the rigid body part of robotic fish, this paper choses the dual-template matchingalgorithm for multi-target tracking and use C++language to implement. Through a series ofexperiments in Visual Studio2008， the results have demonstrated that the targets could be tracked rapidly and accurately in this method, and the real-time performance of the controlsystem will be enhanced remarkably with this algorithm.