The Research On Underwater Object Detection And Tracking Using A Single-Beam Forward Looking Sonar

The technology of underwater target detection and tracking based on sonar is one of the important contents in the domain of Autonomous Underwater Vehicle vision, is the significant technology in aiding navigation and underwater work, is also a precondition of preventing collision from any object.There are two missions in this dissertation: the first, the acoustic visual software system was built based on the embed RTOS VxWorks and pc/104 hardware platform. The module of target detection which takes charge of judging the existence of target in serials of sonar images, describing the shape character of the target was used in the software system. In the second part, the particle filter was studied in serials of the sonar images tracking considering its development trend and tracking precision. The major contents are as follows in general:First, the dissertation summarized the progress trends and current status of research in underwater target detection based on sector-scan sonar, and then introduced the hardware, software and data format of the micron DST sonar system. The next work is the tracking framework which was formed based on VxWorks includes: control orders transmission, data-collecting module, the real-time processing of sonar images data, such as the coordinate transformation, binds of wave inserting, switching of gray level, median filtering, and so on. At the end of the dissertation the particle filter algorithm for tracking forward-looking sonar images was studied. In order to improve the tracking stability, the dissertation adopted the noise-self adapted method which was verified by the serials of sonar images.