| Ime projekta: Uspostava referalne baze za istraživanje utjecaja vremenskih prilika na video nadzor u pomorskom prometu Project’s name: Establishment of reference database for studying the influence of weather conditions on marine video surveillance Akronim / acronym: ERDBSIWCMVS Broj/No.: 2673./2017. Vrijeme / duration: 9.3.2018-9.3.2022. Šifra projekta/Projct code: pfst-2018-03 https://croris.hr/projekti/projekt/5997 |
| Voditelj projekta / Project manager: Igor Vujović |
| Tim/Team: Izv. prof. dr.sc. Igor Vujović (Associate professor) Prof. dr. sc. Danko Kezić (Full professor tenure) Izv. prof. dr. sc. Tatjana Stanivuk (Associate professor) Mr. sc. Ivica Kuzmanić (permanent college professor) Doktorandi: Vladimir Brozović (mentor/supervisor prof. dr. sc. D. Kezić), Bruna Bacelja (mentor/supervisor izv. prof. dr. sc. M. Krčum), Miro Petković (mentor/supervisor izv.porf. dr. sc. Igor Vujović) |
U svijetu postoje referalne baze za prepoznavanje lica, gestikulacija, pokretnih objekata u statičkim scenama, itd. Ne postoji referalna baza za usporedbu rada algoritama u obradi video signala u različitim vremenskim prilikama na istoj sceni. Napredak u istraživanju nadzora prometa u različitim vremenskim uvjetima pridonijet će sigurnosti prometa na kopnu i na vodenim površinama. Detekcija prelaska sektorskih granica bit će olakšana uspostavom ove baze. Uporaba video nadzora iznimno je složena nad vodenim masama, jer postoji niz dodatnih problema koji se ne javljaju u klasičnim nadzornim aplikacijama (detekcija linije horizonta, valovi, maskiranje brzog broda valovima, pjena i kapljice vode, refleksija sunčevog svjetla o vodenu masu, itd.).
There are many referral databases (or datasets) for many purposes in the world, such as for face recognition, gestures, moving objects in static scenes, etc. However, there is no such dataset or base for algorithm performance comparison in video processing, which deals with various weather conditions in the same scene. Advancement in the research of traffic monitoring under various weather conditions will contribute to the traffic safety both on and off shore. Sector borders crossing detection will be easier by the help of this dataset. Video surveillance is especially difficult over great water areas, because there are a lot of additional problems which are specific for maritime zones, and not for land applications (horizon line detection, waves, masking of fast moving object with waves and foam, sea water droplets, sun reflection from the water mass, etc).
| Gospodarski subjekti uključeni u projekt / Corporation out of university community: Tech Resources d.o.o. |
Cilj projekta je stvoriti preduvjete za uspostavu referentne baze te početak uspostave takve baze. Uspostava jednog dostupnog skupa podataka omogućit će znanstvenicima pristup problemu utjecaja vremenskih prilika na segmentaciju kretnji u vodenom okolišu s višom razinom sistematizacije problema.
Želja je da znanstvenici mogu slobodno koristiti ove podatke kako bi mogli razvijati svoje algoritme video nadzora i uspoređivati rezultate s drugim znanstvenicima, koji imaju drugačije algoritme, a proizvoljne dvije video sekvence se ne mogu uspoređivati. Stoga se uspostavom ovakvog skupa podataka ujednačavaju kriteriji za usporedbu rada različitih algoritama koje razvijaju drugi znanstvenici neovisno o zemljopisnom položaju.
The goal of the project is to create a framework for establishment of permanent referent dataset and to start establishing it. Establishment of one available dataset will enable scientists to approach to the complex problem of weather conditions influence to the motion segmentation in maritime zones with higher level of systematization and validation. It is the wish that researches can freely use these dana to develop algorithms in video surveillance field and to compare it with other researches. This project will contribute to equalization of criteria for different algorithms performance comparison no matter where there are around the globe.
Studenti Fakulteta uključeni u projekt / Students involved in the project:
Martina Bradarić, Matea Čović, Bruno Eterović, Nikola Šimunović, Vlado Šušnjar.
Doktorandi koji rade na projektu / PhD candidates working on the project: Miro Petković, Vladimir Brozović, Bruna Bacelja
Referalan baza – skidanje podataka / Benchmark dataset – downloads:
Snimljeni video / Recorded videos: oblačno / cloudy, sunčano /sunny, snježno / snowing vrijeme / weather.
Matlab session in Ground truth labeler with human manual segmentation of recorded videos.
Explanations:
groundTruthLabelingSession_oblacno_ploce.mat – session for cloudy weather
groundTruthLabelingSession_snije_dotjerivanje.mat – session for snowing weather
groundTruthLabelingSession_suncano_ploce.mat – session for sunny weather
You will also need to load labels for cloudy, snowing and/or sunny weather.
Comparison of ground truth determination by three human operators in the same sequence. Txt files contain results of the corresponding jpg files: link to DABAR archive.
Split port zone traffic at different weather conditions – jpg files in zip: 1st part, 2nd part.
| Diplomski radovi iz projekta / Diploma/Master thesis from project: 1. B. E., Prepoznavanje uzoraka iz snimki nadzornog sustava u programskom paketu Labview / Pattern recognition from video surveillance system in Labview, 28. 3. 2018. 2. N. Š., Izdvajanje značajki video signala s ciljem upravljanja kretanjem na zadanoj stazi / Feature extraction from video signal with goal of moving in desired lane, 17.7.2019. Resulting videos (output of the lane detection algorithm): video 1, video 2. 3. V. Š., Proučavanje vidljivosti pomorskih pokretnih objekata u različitim vremenskim uvjetima / Marine traffic visibility in various weather conditions, 30.9.2019. 4. M. B., Video nadzorni sustavi u pomorskom prometu / Video surveillance system sin maritime traffic, 30.9.2019. 5. A. L., Semantička segmentacija kretanja uporabom konvolucijskih neuronskih mreža, 29.9.2020. |
| Published papers |
| Presentation at the ACE-X 2021 conference |
Nabavljena oprema iz projekta / Equipment obtained during the project:
- SSD Samsung 850 EVO Basic 2.5″ 250 GB, MZ-75E250B + dostava x7
- SSD 500.0 GB SAMSUNG 850 EVO Basic, MZ-75E500B, SATA 3, 2.5″, 540/520 MB/s
- Ladica za disk ICY BOX IB-AC649a, SATA 2.5″, za montažu HDD / SSD umjesto optike, 5-9.5 mm
- Digitalna kamera GOPRO HERO7 (Black, 4K60, 12 Mpixela + HDR, Touchscreen, Voice Control, 3 Axis, GPS)
- Kamera bez leća u s pripadajućim USB kabelom i dostavom Lensless camera) MQ022CG-CM s pripadajućim USB kabelom CBL-U3-3M0 i dostavom