Active Binocular Vision Stabilization with a Cerebellum Inspired Model

Active Binocular Vision Stabilization with a Cerebellum Inspired Model

McGuire, K.N.

Jonker, P.P. (mentor)

Mechanical, Maritime and Materials Engineering

BioMechanical Engineering

BMD

2014-12-12

Summary: The need for vision guided mobile robotics is becoming more apparent, however it is proven to be troublesome to stabilize its vision. For two legged robots, it is, during motion, even more difficult than for wheeled robots to be able to perceive the world around it. Finding a stabilization method which is able to reduce the motion blur while walking is essential. Inspiration can be found within the human body itself. The cerebellum is a brain area which is highly involved with stabilization in general, and numerous neurological research has been done by studying stabilizing eye movements. The cerebellum is responsible for adaption within the vestibular ocular reflex (VOR), compensatory eye movements which directly act on signals from the vestibular organ. An active binocular vision based robotic setup will be used to implement this cerebellar model with an inertia sensor, functioning as the “vestibular organ”. During the previous literature survey, several cerebellar models were found to be suitable for implementation, however they do not match the recent neurological discoveries. During this research, two kinds of cerebellar models will be implemented, one existing model and one extended to recent neurological research. Both models will be built using a self learning neural network and will be evaluated by the detected movement within the image. While experimenting with these models, the robotic head will receive both rotational as translational disturbance to simulate the sensation of motion. The cerebellar model will learn to predict the sensory consequence (the optical flow) of the combined control of the inertia sensor and the tracking system and creates a motor input signal which can counter that movement. This results in the robotic head stabilizing its vision better, while it is on the move.

active camera stabilization
cerebellar-inspired model
optical flow

http://resolver.tudelft.nl/uuid:3fb881d2-9701-4f1c-96a6-f6fca74aeec8

2015-01-15

Student theses

master thesis

(c) 2014 McGuire, K.N.