Kovalev V.A., Paulenka D.A., Snezhko E.V., Liauchuk V.A. Comparative analysis of budget computing platforms for a portable micromodule of on-board image classification // BIG DATA and Advanced Analytics: collection of materials of the fourth international scientific and practical conference. (Minsk, Belarus, May 3 – 4, 2018) / editorial board: М. Batura [etc.]. – Minsk, BSUIR, 2018. – pp. 31–42.
Comparative analysis of budget computing platforms for a portable micromodule of on-board image classification
Abstract
This paper is devoted to the analysis of basic hardware and software of recent cheap and commercially
available computing microplatforms for selecting an appropriate solution for development of an onboard micromodule
for preliminary classification and selection of images of underlying surface of given types. It is assumed that the corresponding
versions of the micromodule can be installed on board of small spacecraft or light unmanned aerial vehicles
(drones). In this paper we consider a variant of a micromodule for drones. When choosing a microplatform, the main
limitations were its low weight (no more than 300 grams, including camera and interface equipment) and its relatively
high performance (time for frame processing of a color image 320×240 pixels is no more than 300 milliseconds). Another
important limitation was the low price and commercial availability of micro-platform on the Belarusian market. The
information provided in this paper could be useful for engineers and researchers who develop compact budget mobile
systems for processing, analyzing and classification of images.
Key words: microcomputer, mobile system, image classification, drone, convolutional neural network
Introduction
1. Basic requirements for micromodule being developed.
2. Tested computing microplatforms and their technical specifications.
3. Main software and its features.
4. Results of microplatforms performance testing.
5. Model of micromodule.
6. Composition and general characteristics of the software being developed.
7. A brief SWOT analysis of the tested microplatforms.
8. Conclusions.