Development of two vision estimation algorithms for robot vision control scheme

Abstract

This study deals with the development of two vision estimation algorithms for robot vision control scheme. One is the Extended Kalman Filtering algorithm, and the other is the Newton-Raphson algorithm. The Newton-Raphson (N-R) algorithm consists of vision system model, camera parameters estimation scheme and joint angle estimation scheme. The Extended Kalman Filtering (EKF) algorithm consists of vision system model, process model and measurement model. In addition, the process and the measurement models include the camera parameters estimation scheme and the joint angle estimation scheme, respectively. The vision system model includes six camera internal and external parameters. Each algorithm has its strengths and weaknesses. The Newton-Raphson algorithm is based on iterations and can concurrently handle large amounts of data. On the other hand, it takes a lot of processing time and accordingly is not easy use for real-time robot control. The Extended Kalman Filtering algorithm is based on recursion and thus is faster, but it requires very accurate selection of initial values. In this study we use Monte-Carlo method for estimating initial values. Finally, the results of both algorithms are compared experimentally by tracking the moving target.