Underwater Target Detection And Tracking Research Base On VxWorks

Acoustic detection and tracking system of underwater target is an important component of underwater robot which is a system of intelligence. Exploring new ways for acoustic detection and tracking of underwater target and developing accurate, reliable and real-time detection and tracking software have an important significance to improve the level of intelligent underwater robot.In this paper, we study the sonar image pre-processing techniques in-depth; Extract and analyzed the underwater target characteristics of the sonar image; introduce the information fusion technology into the particle filter theory; implement the underwater target tracking in theoretical framework of particle filter. On this basis, we use SeaKing DST signal-beam imaging sonar and embedded acoustic vision computer as hardware platform and a certain type of underwater robots as the carrier. A detecting and tracking software using real-time multi-tasking operating system-VxWorks was designed, which effectively cooperate with other function modules, to achieve the detecting and tracking of underwater targetIn this paper, the beam interpolation, coordinate transformation, image display, image smoothing and image segmentation pre-processing were completed using sonar data firstly. Then we extracted numerical characteristics of the shape and brightness of underwater targets, shape moments, invariant moment and gray level co-occurrence matrix; compare and analyzed the various characteristic changes of target in motion process; determine the most beneficial combination features for detection and tracking using generalized regression neural network. Final we study the particle filter theory in-depth and find some solutions for its problems. Underwater target tracking was realized in the theoretical framework of particle filter. Tracking process includes initialization, state transition, state observation, re-sampling. Considering the characteristics of the target of sonar images, in order to enhance the effect of particle filter tracking, we substitute the classic model template matching observation model with the information fusion observation model. Two kinds of observation model comparing experimental results show that the information fusion technology into the particle filter can effectively achieve the underwater target tracking. Based on above theoretical research and simulation, we carry out the sonar data reception and communication between other modules of underwater robot in the VxWorks environment; the overall structure of detection and tracking software which includes initialization module, data receiving module, data processing module, TCP module, UDP module are designed.