Velocity Estimation for Autonomous Underwater Vehicles using Vision-Based Systems

Abstract

In this dissertation, it is presented a study of a system architecture capable of calculating the linear and angular velocity of an autonomous underwater vehicle, AUV, in real-time, suitable to be used in the control loop of an AUV. The velocity is estimated using computer vision algorithms, optical flow and block matching, keeping in mind the movement characteristics of autonomous underwater vehicles, i.e. maximum velocity and acceleration, regarding these systems as having a slow dynamic. Considering that these computer vision technics are computing intensive tasks, and are not compatible with real-time systems when implemented in microcomputers, this problem is solved through the study of a possible implementation of these technics in a field programmable gate array, FPGA, and microcomputers. The computer vision algorithms studied, for optical flow computation, were Horn-Schunck, Lucas and Kanade, and it's different variations and optimizations, and more simple algorithms as block matching.