Date of Award

7-14-2016

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Systems Engineering

First Advisor

cang ye

Abstract

The objective of this dissertation is to develop a 3D object recognition method for a portable robotic navigation aid, called “Smart Cane” (SC), for the visually impaired. The SC is a computer-vision-enhanced white cane that uses a single 3D time-of-flight camera for navigation in an indoor environment. To achieve the navigation function, the SC should be able to perform SLAM (Simultaneous Localization And Mapping) and object recognition. In this dissertation, we focus on object recognition. Firstly, a method, named Normal-Coherence-Check RANSAC (RANdom Sample Consensus), is proposed to perform robust and fast plane extraction. Secondly, a 3D object recognition method based on the plane extraction results is proposed. To detect small objects with non-planar surface(s), e.g., doorknob, we introduce a SIFT feature based small object recognition method. It can be used by the SC to detect some types of objects (e. g. a closed door), for which the geometric features based method fail. The method can also be extended for detection of some small objects in office/home environments. Finally, we built a SC prototype to validate the proposed object recognition methods in real-world settings with human subjects. Through the human subject test, we got the real case data by using the SC prototype and we are able to identify the shortcomings, which may guide our future research direction.

Download

Share

COinS