Bestand wählen

Identification of SAR Detected Targets on Sea in Near Real Time Applications for Maritime Surveillance

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Erkannte Entitäten
current thanks ladies gentlemen of next speaker is certainly from the the German Aerospace Center DLR he's gonna to talk to us are
detected targets on the Europe hello everyone thinks for and thanks for listening thanks for 3 today my talk you will be focused on the system for identification and to for detection and edit identification of sigh detected targets that to developed a dealer and to use information so here is a
short outline of my presentation so 1st I want to give you some information about dealer and then the switch to my actual topic you are as a
German Space Agency a so firmly to it's combined to research institutes and has almost 8 thousand employees and to securities concerns such topics like Ultrix space energy and transport it has 16 sites distributed at different locations in Germany and in respect to
remote sensing and GIS there is a receptor Russian center you see which is a cluster of 2 Institute German Remote Sensing Data Center and the Institute for commonsense technologies and to he was he has departments in and is highlighted science so I came from my students there located Department
National ground segment which concerns of data collection from the satellite archiving and management we have several antennas shown in the picture and to and there is a a coverage area by this antennas it's a main research facility for terrorism its mission and now it's also receives a sentinel 1 data and operation iterative so Sosa reduce to and to many other optical and to small science satellite missions and is drawn station also provides a command service for the satellite and here is also located maritime security lapse which uh develops and integrating and real-time services for maritime domain namely we are doing
our target detection from Harold remote sensing satellites hope to going read it includes should detection icebergs and all those pills and also we are doing but I promise you we are deriving the intent state from us senses so about some
detection and identification system so maybe some of you his asking why we needed but uh actually to deviate are highly dependent on the sea notions 90 % of the goods are carried by a uh oceans and torsion produce a lot of food for us 75 millions of on caught fish every year also produce oxygen us major part and safe and secure maritime situation is very important for us so therefore
we created this system which
has already in operation in different projects at the ground station Mr. it operates as a whole receiving sigh emissions listed here goes around and makes it past in the future perhaps a to and Sentinel 1 and 1 a and 2 b of course in the future so as I said become collaborative ground segment and to the data via received a lot of images there so in this case shown in print principle architecture of the system and so on data are being collected from the satellites and then automatically transferred to the processing environment and then processed and system management takes care of the keep it on the fusion of different processes so this is actually a kind of a rapid which schedules different steps have persecution it can around different processors in parallel and wait uh for execution of some others it's a development during development of dealer and uh private company in the variance of the and used in many processing chains and the a lot and here are the main processing steps Richards ecuted for this change so the very initial step after the tangential to the processing system is a level 0 level 1 be process which is sends a dependent for example Florida as a research there is a do a lot developed chemistry processes for example forest sentinel 1 that's the idea of processes and memory some of you would like to reproduce this chain initially if you're not to use ground stations seal would not need this as you will download the data from that say a sentinel scientific up or in the 1 D level the next step is star which transformer this is a pearl software I think does
georeferencing the image projection in different projections it does histogram adjustments and transfer the image to 8 bit format for visualization this process
also and see in other chains and not only in the street it's for example in the ground stations uh at Antarctica it'll used to to and create a clique looks and we can sit and PNG format and that's invite e-mail for fast preview and this is visualization what detection dust on this side is plain and projected image and on the right side and bit it pays you for it is written just 6 so it uses this digital library and uh this uh after we've been doing this feature was which was presented to a deeper recession and 1 room yeah so
the next step is a century but before explain in this step maybe I need to explain what series of for example I'm a newbie been Marina topic and only 2 years ago I had no idea what is this so areas as self-reported identification system so it uses GPS signals and other ship metadata like name and sizes and transmits over the radio channel and all the ships feature appeared to is this device can receive it and which are located in the line of sight so everyone can receive it you can go to the caused by the it's receiver and connection data sometimes you have to infer relatives complaints and there is also a network of base stations which are collecting this information reaches great transmitting them but there also constellation of the satellites which able to receive this and furthermore the next generation of radar satellites perhaps that tourism ix 2 or something else will be created the CAS receiver and then it'll be much easier and much more efficient to do and it emerged probably this is idea all the vessels which which carry more than 300 to 1000 for 12 more helpful for more passengers abbreviated to have this device on board and here is a typical and device a typical screen of the various receiver and the both so there is a mental related shapes so this is a complementary system for collision avoidance and to it complements the on board radar systems sold on
the market there are several that companies which able to provide this information worldwide today uh have access to this base stations and for example acceptors owns constellations of the satellites and to be established to several interfaces which doing a query from From this providers a every providers has its own vision how this data should be stored so different to that dress should be implemented visitors provides and of a feature service which is quite good and easier to handle but some some of them great so fusion and have the fictional structures which union needs to be passed so all these things so this process of and also it can decode format in case of the received our own data and push it into the database for the future you so if you want to reprocess historical data it automatically takes the data from the database and not from external so this is how we
store the data located in or
what happens here in on Microsoft PowerPoint it works itself I didn't mention that
there are a lot of different types of ways messages but there are 2 main types of which are transmitted from the ships 1 contains the voyage information which is dynamical and has actual position uh causal drowned and speed and to some some other attributes and the study confirmation contains actually name of its size and some other identification things and that these 2 types of messages they are transmitted separately and this is a task of analyzes to margin 2 gives a has only 1 key field MMSE number this 1 of his which we can do all this data fusion but unfortunately this enormous number is not really unique we can see the situation that we have achieved as for example in August 0 carbon this 1 intimacy number and 2 the same summer and somehow and this is really a common situation in this is not a problem when the deal with a small a tuple of data from for a small area for small time arranged this we would be obvious it from the same sheet but for the global extended to store this data in the database we have to think about this therefore correct this complex subject and interest and here is
an example it's also to start at
a little bit so the output of this process is adjacent file where ships are just sort of by the moment the number and then inside to be of dynamic and static information and dynamical is sorted by timestamps in people times so it's rated to process performance
of the next step is a should detection itself for that we
have for a saint should detection process this is our main uh processing software it has been developed at another maritime security lap in Bremen the I can truly I give you a lot of you details about this you just integrated in our chain but I know that it is also the you stole the
benefits of 1st rigid tools and libraries for example to use to G. Del to open the libraries for algorithms and tuition detection algorithm based on the far detector with constant false alarm rate which should do analyzes so far arundinaria in both around 2 on certain pixels and can compare the statistical parameters like standard deviation then there is a result of such a binary masks out of reach we can derive a vessel size of lens and it's it's a heating and also In the future tasks guys are working on this and classification and for that we use that as data is a training classes so once we have the
inputs the time to do emerge presenters of another processor is deviation detection evaluator and will which combines the outputs of full previous steps that you referenced image is French and should detection results and so so 0 tasks inside the
1st 1 is that we have to allocate again from they have to reconstruct airstrikes because there's messages are being transmitted every a certain seconds and to is this frequencies depends on the vessel speed but it's really a grand situations that we get exact message of exact image time and especially of set images if you talk about those kinds of the major modes the major at time could take up to 30 seconds and to the anode única timestamp for the image and therefore we have to build to time and depends where the sequence of layers uh on the image we have to use a proper time stamps and then it iteratively adjusted until they get exact points we have to invade at interpolation so therefore hello they're trying to create most tracks and results are brandos and also doing a cross-check on this point falls in the lines 1
another problem is when you do your visit uh Sarah images and moving objects and and uh uh because of traditional Doppler shift the objects appear in the image of in the front positions and has in 1 direction and this shift can reach up to several hundred meters depends on the vessel speed and we have to consider this because otherwise he would do is ranked data fusion and a lot
of different techniques how to procedures this but the most popular is to use some additional hopefully most of the information like traffic enforcement In this case the ASR is traffic and for in all its speed and all but it's direction and then all of the satellite positions can the rest here which will be formally helps to this but if someone is interested i can show it later and so in this the projected messages on the image has a his they appear due to now this Doppler shift and then we are doing data fusion 1st duration is based the nearest neighbor search search and then there are only a doing comparison of estimated parameters to be is uh reported from the areas and to when the differences small we assume that it's a proper object to emerge and I forgot to
mention that everything is based also on an open source libraries like and told you and the last step of of course it's product dissemination have you provide different options as the file uploads to have to be of our work ascending games and files we also serve as on the Jos somewhere there to enable users connect and desktop GIS application also they have mapping gland which is based on you kiss framework which developed also in as a team and to this framework is widely used at the dealer for different projects included in this line and to the front end of this library is based on angular and US library and so
here is an example for our clients here's overlaid mentioned on the right side have the scenes from different
sensors service different value layers like motion detection in 2 waves and to case where Saul here we overlay event detection product which looks like this into the query feature for was way for future for you've used to a templates on of somewhere which allows us to provide a unified
efficient for responders independently of the use of however might inclined toward the use of a disco publications will show you later and in this example relates the detection should
detection results with serious messages is sufficient for quite often I was asked about how we embed this someone else and their future and for requests that response do that that visit generates for every detection of small wasn't in picture as some male and then push them and use over and the image was
I but module allows us to create a customer and dimensions and then we can
put the can push several hundreds of something else but store them in just over as 1 single data store 1 single layer and then have access to know this such simple query where we defined a gene you won't and understand you
again have and desktop
applications like bodies were dubious and to the
same view recreated in our candidates products have put their cuckoo he mentioned all relate these areas extension detections but in Durkheim's that file also put but the
tracks information and to user can do animation how this stress moment in time so feel kind the
details here for objects in the morning in different directions and it was data fusion is done and if you look on the the tubules so this
is our detection and this is serious so we see that derived parameters are very close to reported for example the estimated average convinced 319 by 40 meters and reported 299 by 48 which is quite good it's results are a green the same this and he did while you FIL difference on 6 degrees which is acceptable and if you do a query about all the fun Marina traffic website which is open everyone can do a query with is known as the number you
see what's shippers this cell and
she this feature here and this
kind of training on what is this all about to do a
conclusion so we developed a to a system which works for a near real-time applications we provide results as in 10 15 minutes after the image creation from the satellite and did useful in several projects fully automated and simple logic behind dispute out of open source libraries such as g dull all would you tend to have Joseph and and so on this is an example of how it can work and the last very last but not least I mentioned that of work this optical
sensors and here's another framework but it probably a speech for NASA
conference that's my the next year it's also based on false reject those and it has also we we stumbled into API from this is
how it looks like of them down
thanks to work that thank and questions the could repeat what you
said about the Doppler effect was a good reader or also yesterday this radar images and the target is moving towards the satellite the where the sensor at this moment in time due introduces additional Doppler shift and as a result the objects are shifted from the image from its real position and this actually this effect is used to form in a speech music music the traffic patrols and to by 1st announced to the press reach measuring the the speed of movie stars as well this the question but the to say that the bits of eraser bills with lots of open source tools of any bits of the the kind of the tools you develops other open-source themselves can we get the use the tools of cells as the moments now it's a really uh developed only for our system but maybe in the future we will publish some components for example metadata parsers for address satellites which can be integrated and you know that that's an idea maybe for the future development the question that I have a question and can use the same detection tool for same measuring the velocity of the when yes yes but it would be difficult to yeah there are some techniques this refocusing the the mentioned we can extract actual position and out of this is the speed uh but it's it's quite difficult that without additional information about possible and there already was some studies before it's the same objects the same woman targets much it perhaps it would be not that accurate is this alright well the questions of the romances and
Service provider
Gruppe <Mathematik>
Varietät <Mathematik>
Generator <Informatik>
Kollaboration <Informatik>
GRASS <Programm>
Einheit <Mathematik>
Objekt <Kategorie>
Folge <Mathematik>
Elektronische Bibliothek
RADAR <Automatisierungssystem>
Mathematische Logik
Weg <Topologie>
Endogene Variable
Unified Threat Management
Prozess <Physik>
Inferenz <Künstliche Intelligenz>
Weg <Topologie>
Komplex <Algebra>
Maschinelles Sehen
Algorithmische Programmiersprache
Konfiguration <Informatik>
Spannweite <Stochastik>
Framework <Informatik>
Projektive Ebene
Web Site
Zellularer Automat
Physikalisches System
Speicher <Informatik>
Operations Research
Bildgebendes Verfahren
Displacement Mapping
Einfach zusammenhängender Raum
Wald <Graphentheorie>
Physikalisches System
Objekt <Kategorie>
Desintegration <Mathematik>
Kartesische Koordinaten
Shape <Informatik>
Mosaicing <Bildverarbeitung>
Güte der Anpassung
Dienst <Informatik>
Rechter Winkel
Mathematische Logik
Klasse <Mathematik>
Dienst <Informatik>
Service provider
Verband <Mathematik>
Attributierte Grammatik
Open Source
Elektronische Publikation
Eigentliche Abbildung
Kartesische Koordinaten
Funktion <Mathematik>
Nichtlinearer Operator
Message sequence chart
Prozess <Informatik>
Verkettung <Informatik>
Funktion <Mathematik>
Technische Optik
Reelle Zahl
Google Earth
Kombinatorische Gruppentheorie
Framework <Informatik>
Fluss <Mathematik>
Zusammenhängender Graph
Diskretes System
Einfache Genauigkeit
Mapping <Computergraphik>
Shape <Informatik>


Formale Metadaten

Titel Identification of SAR Detected Targets on Sea in Near Real Time Applications for Maritime Surveillance
Serientitel FOSS4G Bonn 2016
Teil 107
Anzahl der Teile 193
Autor Voinov, Sergey
Schwarz, Egbert
Krause, Detmar
Berg, Matthias
Lizenz CC-Namensnennung 3.0 Deutschland:
Sie dürfen das Werk bzw. den Inhalt zu jedem legalen Zweck nutzen, verändern und in unveränderter oder veränderter Form vervielfältigen, verbreiten und öffentlich zugänglich machen, sofern Sie den Namen des Autors/Rechteinhabers in der von ihm festgelegten Weise nennen.
DOI 10.5446/20456
Herausgeber FOSS4G, Open Source Geospatial Foundation (OSGeo)
Erscheinungsjahr 2016
Sprache Englisch

Inhaltliche Metadaten

Fachgebiet Informatik
Abstract Remote sensing technologies are widely used in maritime surveillance applications. Nowadays, spaceborne Synthetic Aperture Radar (SAR) systems provide outstanding capabilities for target detection on sea for large areas independently from the weather conditions. Complementary information from the Automatic Identification System (AIS) makes it possible to generate a value-added target-detection product. Resulting layers on the maritime situation – once provided to decision makers - would be highly beneficial in order to understand what is going on at sea and how it would impact on Maritime Safety and Security. This paper describes the approach of SAR-AIS data fusion and its visualization means developed for Near Real Time (NRT) Applications for Maritime Situational Awareness.

Ähnliche Filme