A Computer Vision Encyclopedia-Based Framework with Illustrative UAV Applications

Terence Morley,Martin Turner,Tim Morris

doi:10.3390/computers10030029

Terence Morley, Martin Turner + Show 1 more

Open Access

https://doi.org/10.3390/computers10030029

Copy DOI

Abstract

This paper presents the structure of an encyclopedia-based framework (EbF) in which to develop computer vision systems that incorporate the principles of agile development with focussed knowledge-enhancing information. The novelty of the EbF is that it specifies both the use of drop-in modules, to enable the speedy implementation and modification of systems by the operator, and it incorporates knowledge of the input image-capture devices and presentation preferences. This means that the system includes automated parameter selection and operator advice and guidance. Central to this knowledge-enhanced framework is an encyclopedia that is used to store all information pertaining to the current system operation and can be used by all of the imaging modules and computational runtime components. This ensures that they can adapt to changes within the system or its environment. We demonstrate the implementation of this system over three use cases in computer vision for unmanned aerial vehicles (UAV) showing how it is easy to control and set up by novice operators utilising simple computational wrapper scripts.

Highlights

Unmanned aerial vehicles (UAV) are being increasingly used as input directly for specific imaging applications [1,2,3]
We refer to small, low-altitude, commercial off-the-shelf (COTS) UAVs as these have the advantage of mass availability and low cost and are already being exploited by major organisations and in many and various imaging applications
The encyclopedia contains parameters for various functions in the algorithm; meta parameters to calculate parameters dependent on the particular circumstances; and input data which may come from input sensors or from the output of other pipelines running within the framework

Summary

Introduction

Unmanned aerial vehicles (UAV) are being increasingly used as input directly for specific imaging applications [1,2,3]. The authors are currently performing research for a UK company operating in the maritime sector Their particular problem has a similar scenario, in that UAVs will be searching for sea-borne craft, in coastal areas. In the event that the real-time UAV monitoring system failed to identify a specific maritime object, it would be of future benefit to be able to re-analyse the collected video streams with different image processing algorithms involving different computational resources. Rather than paying for even a smaller inter-operable system, a company can buy software modules that will cross-interoperate This may require some system integration by the end-customer but would bring the added benefit of agility to their system. We envisage the operation of this framework to be similar to a smart mobile phone operating system and infrastructure model where users can select the appropriate app without needing to know anything specific about particular system integration

Framework

Use Cases

Use Case 1

Findings

Use Case 2