VLSI architecture-based implementation of motion estimation algorithm for  Underwater Robot Vision System

Aarti Hemant Tirmare; Priyadarshani Shivakumar Mali; Amardeep Anandrao Shirolkar; Ganesh Rajaram Shinde; Vikas Dattatray Patil; Hemant Appa Tirmare

doi:10.31838/jvcs/06.02.13

Authors

Aarti Hemant Tirmare Assistant Professor, Department of Electronics And Telecommunication Engineering, Bharati Vidyapeeth's college of Engineering, Kolhapur, Maharashtra
Priyadarshani Shivakumar Mali Assistant Professor, Department of Electronics And Telecommunication Engineering, Bharati Vidyapeeth's college of Engineering, Kolhapur, Maharashtra
Amardeep Anandrao Shirolkar Assistant professor, Electronics and Telecommunication Engineering Department of Technology, Shivaji University, Kolhapur, Maharashtra
Ganesh Rajaram Shinde Assistant Professor, Mechanical Engineering , Department of Technology, Shivaji University, Kolhapur, Maharashtra
Vikas Dattatray Patil Assistant Professor, Department of Electronics And Telecommunication Engineering, Bharati Vidyapeeth's college of Engineering, Kolhapur, Maharashtra
Hemant Appa Tirmare Assistant Professor, Computer Science and Technology , Department of Technology, Shivaji University, Kolhapur, Maharashtra

DOI:

https://doi.org/10.31838/jvcs/06.02.13

Keywords:

Motion Estimation, Underwater Robot Vision, Computer vision, FPGA, Velocity.

Abstract

Motion Estimation (ME) is a computationally intricate issue often affected by non-exact solutions. Consequently, it is imperative to embrace many methodologies and assumptions. Velocity is an essential variable for Underwater Robot Vision (URV) navigation, yet its
estimation poses significant challenges. Conventional techniques, such as the Global Positioning System (GPS), are ineffective underwater, while alternative approaches need to calculateocean current's velocity or employ sound sensors. Nonetheless, these systems possess inherent limits. A viable method to calculate the velocity of URVs is through the utilization of Computer Vision (CV) systems when a marine structure is inside the URVs field of view, as these sensors are not affected by the same issues as acoustic routing or the influence of ocean currents. This study devised an algorithm that estimates a vehicle's rotational velocity (RV) and axial velocity (AV) by utilizing the motion field derived from the optical ME technique. The algorithms were evaluated in simulated settings, and the findings were shown using several approaches for ME. The ME method was found to function as predicted, with a maximum error of 2.23% in the conducted tests, and showed greater robustness under challenging underwater situations. This work also presented a hardware Programmable Gate Arrays (FPGAs) to enable real-time computation of dense optical motion for Video Graphics Array (VGA) images, utilizing the ME method.