mdh.sePublications
Change search
Link to record
Permanent link

Direct link
BETA
Ahlberg, Carl
Publications (4 of 4) Show all publications
Ahlberg, C. (2020). Embedded high-resolution stereo-vision of high frame-rate and low latency through FPGA-acceleration. (Doctoral dissertation). Västerås: Mälardalen University
Open this publication in new window or tab >>Embedded high-resolution stereo-vision of high frame-rate and low latency through FPGA-acceleration
2020 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

Autonomous agents rely on information from the surrounding environment to act upon. In the array of sensors available, the image sensor is perhaps the most versatile, allowing for detection of colour, size, shape, and depth. For the latter, in a dynamic environment, assuming no a priori knowledge, stereo vision is a commonly adopted technique. How to interpret images, and extract relevant information, is referred to as computer vision. Computer vision, and specifically stereo-vision algorithms, are complex and computationally expensive, already considering a single stereo pair, with results that are, in terms of accuracy, qualitatively difficult to compare. Adding to the challenge is a continuous stream of images, of a high frame rate, and the race of ever increasing image resolutions. In the context of autonomous agents, considerations regarding real-time requirements, embedded/resource limited processing platforms, power consumption, and physical size, further add up to an unarguably challenging problem.

This thesis aims to achieve embedded high-resolution stereo-vision of high frame-rate and low latency, by approaching the problem from two different angles, hardware and algorithmic development, in a symbiotic relationship. The first contributions of the thesis are the GIMME and GIMME2 embedded vision platforms, which offer hardware accelerated processing through FGPAs, specifically targeting stereo vision, contrary to available COTS systems at the time. The second contribution, toward stereo vision algorithms, is twofold. Firstly, the problem of scalability and the associated disparity range is addressed by proposing a segment-based stereo algorithm. In segment space, matching is independent of image scale, and similarly, disparity range is measured in terms of segments, indicating relatively few hypotheses to cover the entire range of the scene. Secondly, more in line with the conventional stereo correspondence for FPGAs, the Census Transform (CT) has been identified as a recurring cost metric. This thesis proposes an optimisation of the CT through a Genetic Algorithm (GA) - the Genetic Algorithm Census Transform (GACT). The GACT shows promising results for benchmark datasets, compared to established CT methods, while being resource efficient.

Abstract [sv]

Autonoma agenter är beroende av information från den omgivande miljön för att agera. I en mängd av tillgängliga sensorer är troligtvis bildsensorn den mest mångsidiga, då den möjliggör särskillnad av färg, storlek, form och djup. För det sistnämnda är, i en dynamisk miljö utan krav på förkunskaper, stereovision en vanligt tillämpad teknik. Tolkning av bildinnehåll och extrahering av relevant information går under benämningen datorseende. Datorseende, och specifikt stereoalgoritmer, är redan för ett enskilt bildpar komplexa och beräkningsmässigt kostsamma, och ger resultat som, i termer av noggrannhet, är kvalitativt svåra att jämföra. Problematiken utökas vidare av kontinuerlig ström av bilder, med allt högre bildfrekvens och upplösning. För autonoma agenter krävs dessutom överväganden vad gäller realtidskrav, inbyggda system/resursbegränsade beräkningsplattformar, strömförbrukning och fysisk storlek, vilket summerar till ett otvetydigt utmanande problem.

Den här avhandlingen syftar till att åstadkomma högupplöst stereovision med hög uppdateringsfrekvens och låg latens på inbyggda system. Genom att närma sig problemet från två olika vinklar, hårdvaru- och algoritmmässigt, kan ett symbiotiskt förhållande däremellan säkerställas.Avhandlingens första bidrag är GIMME och GIMME2 inbyggda visionsplattformar, som erbjuder FPGA-baserad hårdvaruaccelerering, med särskilt fokus på stereoseende, i kontrast till för tidpunkten kommersiellt tillgängliga system.Det andra bidraget, härrörande stereoalgoritmer, är tudelat.Först hanteras skalbarhetproblemet, sammankopplat med disparitetsomfånget, genom att föreslå en segmentbaserad stereoalgoritm.I segmentrymden är matchningen oberoende av bildupplösningen, samt att disparitetsomfånget definieras i termer av segment, vilket antyder att relativt få hypoteser behövs för att omfatta hela scenen.I det andra bidraget på algoritmnivå, mer i linje med konventionella stereoalgoritmer för FPGAer, har Censustransformen (CT) identifierats som ett återkommande kostnadsmått för likhet. Här föreslås en optimering av CT genom att tillämpa genetisk algoritm (GA) - Genetisk Algoritm Census Transform (GACT). GACT visar lovande resultat för referensdataset jämfört med etablerade CT-metoder, men är samtidigt resurseffektiv.

Place, publisher, year, edition, pages
Västerås: Mälardalen University, 2020
Series
Mälardalen University Press Dissertations, ISSN 1651-4238 ; 304
Keywords
Computer vision, stereo vision, FPGA, embedded systems
National Category
Computer Vision and Robotics (Autonomous Systems)
Research subject
Computer Science
Identifiers
urn:nbn:se:mdh:diva-46240 (URN)978-91-7485-453-4 (ISBN)
Public defence
2020-01-28, Kappa, Mälardalens högskola, Västerås, 09:15 (English)
Opponent
Supervisors
Available from: 2019-12-04 Created: 2019-12-04 Last updated: 2020-01-10Bibliographically approved
Ahlberg, C., Leon, M., Ekstrand, F. & Ekström, M. (2019). Unbounded Sparse Census Transform using Genetic Algorithm. In: 2019 IEEE WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV): . Paper presented at 19th IEEE Winter Conference on Applications of Computer Vision (WACV), JAN 07-11, 2019, Waikoloa Village, HI (pp. 1616-1625). IEEE
Open this publication in new window or tab >>Unbounded Sparse Census Transform using Genetic Algorithm
2019 (English)In: 2019 IEEE WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV), IEEE , 2019, p. 1616-1625Conference paper, Published paper (Refereed)
Abstract [en]

The Census Transform (CT) is a well proven method for stereo vision that provides robust matching, with respect to object boundaries, outliers and radiometric distortion, at a low computational cost. Recent CT methods propose patterns for pixel comparison and sparsity, to increase matching accuracy and reduce resource requirements. However, these methods are bounded with respect to symmetry and/or edge length. In this paper, a Genetic algorithm (GA) is applied to find a new and powerful CT method. The proposed method, Genetic Algorithm Census Transform (GACT), is compared with the established CT methods, showing better results for benchmarking datasets. Additional experiments have been performed to study the search space and the correlation between training and evaluation data.

Place, publisher, year, edition, pages
IEEE, 2019
Series
IEEE Winter Conference on Applications of Computer Vision, ISSN 2472-6737
National Category
Computer and Information Sciences
Identifiers
urn:nbn:se:mdh:diva-44332 (URN)10.1109/WACV.2019.00177 (DOI)000469423400170 ()2-s2.0-85063571752 (Scopus ID)978-1-7281-1975-5 (ISBN)
Conference
19th IEEE Winter Conference on Applications of Computer Vision (WACV), JAN 07-11, 2019, Waikoloa Village, HI
Available from: 2019-06-20 Created: 2019-06-20 Last updated: 2019-12-18Bibliographically approved
Loni, M., Ahlberg, C., Daneshtalab, M., Ekström, M. & Sjödin, M. (2018). Embedded Acceleration of Image Classification Applications for Stereo Vision Systems. In: Design, Automation & Test in Europe Conference & Exhibition DATE'18: . Paper presented at Design, Automation & Test in Europe Conference & Exhibition DATE'18, 19 Mar 2018, Dresden, Germany.
Open this publication in new window or tab >>Embedded Acceleration of Image Classification Applications for Stereo Vision Systems
Show others...
2018 (English)In: Design, Automation & Test in Europe Conference & Exhibition DATE'18, 2018Conference paper, Published paper (Other academic)
National Category
Engineering and Technology Computer Systems
Identifiers
urn:nbn:se:mdh:diva-40877 (URN)
Conference
Design, Automation & Test in Europe Conference & Exhibition DATE'18, 19 Mar 2018, Dresden, Germany
Projects
DPAC - Dependable Platforms for Autonomous systems and ControlDeepMaker: Deep Learning Accelerator on Commercial Programmable Devices
Available from: 2018-09-20 Created: 2018-09-20 Last updated: 2018-09-20Bibliographically approved
Ahlberg, C., Leon, M., Ekstrand, F. & Ekström, M.The Genetic Algorithm Census Transform.
Open this publication in new window or tab >>The Genetic Algorithm Census Transform
(English)Manuscript (preprint) (Other academic)
National Category
Embedded Systems
Identifiers
urn:nbn:se:mdh:diva-46244 (URN)
Available from: 2019-12-04 Created: 2019-12-04 Last updated: 2019-12-04Bibliographically approved
Organisations

Search in DiVA

Show all publications