Overview

The Stingray Project is an unmanned underwater vehicle (UUV) or autonomous underwater vehicle (AUV) specifically designed with the AUVSI AUV Competition in mind. The vehicle currently does all of its navigation based on vision and object detection combined with readings from a Inertial Navigation System (INS). The focus of my research will be in underwater navigation techniques, including vision based navigation and underwater positioning techniques. Due to sensors currently available to me for research and the cost prohibitive nature of acquiring sensors for better positiioning techniques, my research will be primarily the uses of vision for navigation. My work in vision will be based directly on the AUVSI AUV Competition. This is a valid starting point for this research since the competition itself was designed with the future use of such underwater vehicles as a foundation. The Stingray Wiki contains details about the project, including how to get involved, download the code, and current tasking.

Updated on April 29, 2010

Methods

GENERAL

  • Techniques to Consider:
    • Pre-processing Image for Cloudy Water
    • Video Bytes - FPGA Preprocessing
    • Boosting for Best Classifiers
      • JBoost has great documentation for setup
      • Using the -c filename option creates a C file that can do the boost tree on a given pixel's data

BUOY

PIPE

Updated on April 01, 2010

Important Notes

GENERAL

  • The RGB color space has a weakness in representing shading effects or rapid illumination changes. [5]
  • Process ADT by accumulating the count while traversing the tree (BFS/DFS), starting with the value given at the root of the tree. [6]

Updated on October 27, 2009

Hardware

The Stingray vehicle uses a lot of different hardware and sensors. Listed here are those sensor that will be directly accessed for use in this research as it is currently defined.

  • 2 Cameras (NTSC)
  • Frame Grabber - VFG7330
    • Size: 3.6 x 3.8 x 0.6 inches (90 x 96 x 15 mm)
    • Weight: 0.14 lbs (0.06 Kg)
    • Inputs: 2 Channel accepting (NTSC, PAL, SECAM)
    • Output: 640X480 RGB Image Array
    • Speed: Up to 30fps
  • 1 Shared Processor

This research would benefit for better and new sensors and so would the vehicle in general. Listed here are those sensors that could improve the research as it is currently defined, as well as sensors that would allow the research to expand by incorporating other techniques for a general positioning capability.

  • 2 New Cameras
  • 1 Dedicated Vision Processor
  • 1 Velocity Sensor
  • 1 New Intertial Navigation System (INS)

Updated on November 03, 2009

Software

The Stingray software is written in C and C++ and runs in a Ubuntu-based Linux distribution with the Real-Time Application Interface (RTAI) built into the kernel. The main components of the software are planner, vision, and navigation. The planner controls higher level concepts such as direction through the navigation unit or vision mode through the vision unit. There is also a GUI called the Dock Control Station (DCS), which is used to control all aspects of the vehicle when tethered.

This research will be done primarily in the vision component of the system. In addition some work may be done in the other components to help with productivity (DCS) or to utilize new break-throughs in the vision capabilities (NAV). Potential software development could be done on stand alone applications to test and demonstrate potential image processing techniques. However, wherever possible, this will be rolled into the current Stingray software package.

Updated on September 17, 2009

References

  • [1] Leonard, John J. and Bennett, Andrew A. and Smith, Christopher M. and Jacob, Hans and Feder, S. Autonomous Underwater Vehicle Navigation. MIT Marine Robotics Laboratory Technical Memorandum (1998).
  • [2] Balasuriya, B.A.A.P. and Takai, M. and Lam, W.C. and Ura, T. and Kuroda, Y. Vision based autonomous underwater vehicle navigation: underwatercable tracking. OCEANS '97. MTS/IEEE Conference Proceedings (1997).
  • [3] Garcia, R. and Batlle, J. and Cufi, X. and Amat, J. Positioning an underwater vehicle through image mosaicking. Robotics and Automation (2001). Proceedings IEEE International Conference (2001).
  • [4] Viola, Paul and Jones, Michael. Robust Real-time Object Detection. International Workshop on Statistical and Computational Theories of Vision - Modeling, Learning, Computing, and Sampling (2001).
  • [5] T. Darrell, G. Gordon, M. Harville, J. Woodfill, "Integrated Person Tracking Using Stereo, Color, and Pattern Detection," In Proceedings of IEEE Computer Vision and Pattern Recognition, pp. 601-608, June 1998.
  • [6] Fruend, Yoav. The alternating decision tree learning algorithm. In Machine Learning: Proceedings of the Sixteenth International Conference (1999).

Updated on October 29, 2009