Title

Discover Vision: A Framework For Building, Evaluating, And Testing Performance Based Machine Vision Applications

Keywords

Autonomous vehicles; Context-based reasoning; Machine learning; Machine vision; Testing and simulation; Unmanned vehicles

Abstract

This paper presents Discover Vision, a framework for the fast creation, evaluation, and testing of machine vision applications used in real time systems such as autonomous vehicles. The framework utilizes user edited scripts describing what image processing and feature extraction techniques to employ. Users can easily and quickly build a vision system by altering these scripts without changing the underlying framework, thus saving time when testing new methods. A graphical user interface is used to display the real time performance in the form of visual displays, processed frame rates, and system accuracy based on validation sets. It is possible to evaluate the effectiveness of a script by loading in live or recorded video for visual or numerical analysis. Scripts developed in Discover Vision can be used within a custom framework through the use of the scripting engine. Extensibility is achieved through a plug-in architecture. This paper describes the Discover Vision framework, demonstrates its application in an autonomous ground vehicle, and analyzes the resulting performance.

Publication Date

1-1-2006

Publication Title

AUVSI Unmanned Systems North America Conference 2006

Volume

1

Number of Pages

335-347

Document Type

Article; Proceedings Paper

Personal Identifier

scopus

Socpus ID

84897532096 (Scopus)

Source API URL

https://api.elsevier.com/content/abstract/scopus_id/84897532096

This document is currently not available here.

Share

COinS