Underwater Target Detection And Tracking Research Base On VxWorks

Underwater Target Detection and Tracking System is theeye ofAUV.It is a basic part of underwater robot which can guarantee underwater robot autonomous underwater activities.An accurate and reliable,real-time Detection and Tracking System plays an important role of the underwater work.It has practical significance to promote ocean development and protection too.This paper focuses on the sonar image processing,feature extraction of target,target tracking technology,knowledge of real-time multitasking VxWorks operating system.Designed a high reliability and real-time sound detection and tracking system.This system allows underwater robots conjunct with other modulesof underwater autonomous robot to complete a variety of complex taskas a whole.The system mainly includes four parts: the sonar image preprocessing,moving target detection and feature extraction,target tracking,hardware platforms and software architecture of detection and tracking system.Firstly,system handles received echo datafrom sonar.Display the image by programming.Then through coordinate transformation,image smoothing and median filtering processes to achieve the sonar image preprocessing,it is ready for subsequent target detection.Secondly,in moving target detection and feature extraction procedure,the system combine clustering segmentation and variable frame scanning as a method to detect moving targets.Then compared the characteristics of sonar image targets and the changes in target features.Analysis the moment invariant features of the target.Asonar image feature extraction method focused on a new class-based Haar features are designed.The experimental analysis and results show that the effectof image feature extraction is improved obviously by this method.Thirdly,in tracking section,since the target is moving,the feature of target is unstable and the shape of the target changes a lot.Based on the fact that the shape of the target in the sonar image was changing in every frame,compared the method combined particle filter and deformable template with the method combined particle filter andMulti feature information fusionwas proposed and proved by experiments,And select the second methods.Finally,in the acoustic detection tracking software architecture section,I designed serial data receiver module in Tornado cross-development environment and used VxWorks' smultithreading,real-time multi-tasking capabilities,network communication function,then developed a data processing module,I also used the socket network programming interface to achieve a two-way communication between the master and this system.