58: AI, muons, terahertz, smugglers beware! New technologies are here
Duration 00:29:20
Apple Podcasts
Spotify
YouTube
Transcript
CORDIScovery podcast- Episode #58 - AI, muons, terahertz, smugglers beware! New technologies are here.
This is an AI transcription.
00:00:15:16 - 00:00:48:20
Abigail Acton
Hello and welcome to this episode of CORDiscovery with me, Abigail Acton. Increasing amounts of dangerous and illegal substances pass through the EU, as the number of parcels being sent mount and shipping systems get more porous. According to the European Commission, the volume of small packages coming into the EU has doubled every year since 2022. In 2024 alone, 4.6 billion small packages entered the EU market, and this is not including the number of packages shipped within the EU itself.
00:00:49:00 - 00:01:21:12
Abigail Acton
How things are shipped is changing. These increasing volumes are now passing through postal agencies, private vehicles and small logistics companies. The analysis and detection capabilities in such locations are very low, so the process is a target for the smuggling of illicit goods. When it comes to bigger container shipments, ports are also a target. The European Ports Alliance gathers together operators, shipping associations, customs and law enforcement authorities to combat organised crime and drug trafficking by improving security imports.
00:01:21:12 - 00:01:34:11
Abigail Acton
But all advances in technology are very welcome. So how can the work being done by our three guests, all of whom have received support from the EU's research and innovation funding, have an impact?
00:01:34:13 - 00:01:51:09
Abigail Acton
Siiri Salupere is a research fellow in radiation protection at the University of Tartu in Estonia. Her work focuses on ionising radiation, looking at it as a potential hazard to human health, and also exploring the potential benefits of using cosmic rays for security applications. Welcome, Siiri.
00:01:51:11 - 00:01:52:23
Siiri Salupere
Hello. Great to be here.
00:01:53:00 - 00:02:08:13
Abigail Acton
Lovely to have you. Tarvo Metspalu is a product development engineer at GScan in Estonia, where he reviews hardware designs and machine learning data pipelines. He has been working on developing a more portable and efficient analysis and detection device for customs. Hi Tarvo!
00:02:08:15 - 00:02:10:14
Tarvo Metspalu
Hello. Thank you for inviting me.
00:02:10:15 - 00:02:35:08
Abigail Acton
Thank you for coming. Marco Paleari holds a PhD in effective computing from Telecom ParisTech, where he specialised in the intersection of AI and human emotions. He is currently a research and innovation project manager at the ‘AI data for digital defence’ unit at ENG, Italy. Marco describes himself as someone whose interests spanned just about everything, provided it contains a computer chip. Hi Marco.
00:02:35:12 - 00:02:37:11
Marco Paleari
Hello, Abigail. Glad to be here.
00:02:37:12 - 00:03:06:13
Abigail Acton
Very welcome. Siiri, I'm going to turn to you first, please. The SilentBorder project is developing a new high tech cosmic ray tomography, CRT for short, scanner that enables fast and safe screening, detection and identification of hazardous and illegal goods, contraband, and people hiding in containers. We're going to be talking a lot about muons today series. So can we just start by establishing what they are, what properties they have that make them useful as detectors?
00:03:06:15 - 00:03:45:22
Siiri Salupere
The muons are a part of cosmic radiation that is always surrounding us, and we can call muons the secondary cosmic radiation as they are produced in Earth's atmosphere when high energy particles from out of our planet bombard atmospheric particles, and then a chain of nuclear reactions occur and nuons are formed, they are actually elementary particles and quite similar to electrons, but much higher in mass.
00:03:45:22 - 00:03:54:01
Siiri Salupere
And that makes them distinct to electrons. They can penetrate more material than the electrons too.
00:03:54:02 - 00:03:55:08
Abigail Acton
Because they're bigger.
00:03:55:12 - 00:03:56:23
Siiri Salupere
Because they are bigger.
00:03:57:04 - 00:03:57:18
Abigail Acton
More massive.
00:03:57:18 - 00:04:30:24
Siiri Salupere
And they are more inert, so they can't be absorbed that quickly. And on the other hand, we could compare them to X-rays. And, in a more wider sense, we would call it electromagnetic radiation. Now, if we compare muons to electromagnetic magnetic radiation, the big change comes because they have a charge, which means that they scatter more than the X-rays would do.
00:04:31:00 - 00:04:38:11
Abigail Acton
So the fact that they have a charge, an electrical charge, means that when they impact on something, they scatter more?
00:04:38:14 - 00:05:05:19
Siiri Salupere
Yes. And now this scattering effect is dependent on the material. The higher the atomic mass of the material, the more muons are scattering, because there is an electromagnetic force between the positive charge of the nucleus of the matter and the negative charge of the muon.
00:05:05:20 - 00:05:25:10
Abigail Acton
So the muons are much more reactive when they come into contact with, when they hit, an object that's perhaps hidden or whatever, and the manner in which they react and how they scatter can give you information about the nature of the object that they're hitting. Is that correct?
00:05:25:11 - 00:05:43:10
Siiri Salupere
Exactly. And then if for X-rays, we could only distinguish material based on their density, then muons carry the information about the density of the material and also about the material composition.
00:05:43:10 - 00:05:52:09
Abigail Acton
Because they scatter in a certain way each time the composition changes. So if you can trace the patterns of the scattering, you have an idea of the composition of the material?
00:05:52:14 - 00:06:03:02
Siiri Salupere
Yes. And this can open up a window of being able to detect light materials which remain unseen for the common X-rays.
00:06:03:06 - 00:06:22:03
Abigail Acton
Okay, this is a fantastic explanation. Thank you so much. It's much clearer because, you know, we hear the word, but we don't necessarily know what it means or how it can be used practically. So now we've established a little bit about what muons are and why they're of interest to people who are trying to use them for the detection of materials in packages or in containers.
00:06:22:04 - 00:06:27:04
Abigail Acton
Can you then tell me a little bit about how your project harnessed this capacity?
00:06:27:06 - 00:07:02:12
Siiri Salupere
If we build a detector that has the possibility to detect the muon, and we have one detector plate on top of the object and another one underneath the object, then we are able to understand how the muon has been travelling through the matter, by the fact that we can understand how it has been scattering in the matter, and this enables us to detect what has been in the inside, the object that we have been scanning.
00:07:02:13 - 00:07:09:01
Abigail Acton
So this is great. So this is the technical capacity of muons. How did SilentBorder harness that?
00:07:09:03 - 00:07:47:02
Siiri Salupere
Yeah. So these objects inside the cargo we scan can possibly be something illegal, something that's not licensed. And why not even people. So this technology can help us to distinguish light elements. It can be something powdery. It can be a narcotic substance. And the current X-ray would not be able to see these differences because of the different properties of the X-ray and muons.
00:07:47:06 - 00:08:12:24
Siiri Salupere
And another fun fact about the muon technology is the fact that we are using a natural flux of radiation, meaning that we do not create any extra radiation dose which X-rays would produce. So even if there is a person hidden inside this cargo, they would not be getting any harmful effects from ionising radiation.
00:08:12:24 - 00:08:20:19
Abigail Acton
Right. So in the SilentBorder project, did you then construct a prototype to show how this could be used?
00:08:20:21 - 00:08:52:05
Siiri Salupere
Yes. The SilentBorder project built the prototype instrument of the scanner, and we also tried it out in real world trying to see the difference in light powdery materials like flour, starch, sugar, salt. And the technology really showed that it's possible to make the difference. All the building work was actually done by GScan, so you will hear more about GScan when Tarvo will be speaking.
00:08:52:05 - 00:08:56:13
Siiri Salupere
And our object to be scanned was a sea container.
00:08:56:13 - 00:09:07:02
Abigail Acton
So something very very big. So excellent. So how does this kind of work? You have a plate underneath and a plate above and you manipulate the container between the two like a giant sandwich? Or how does that go?
00:09:07:04 - 00:09:53:16
Siiri Salupere
Basically it is like a giant arch and the scanner is scanning what is underneath it, and the container is slowly passing through this scanner. Actually, this scanning takes quite a lot of time. It's 50 minutes, so we are not competing with the X-ray technology, but the SilentBorder technology could be a useful complementary technology to X-rays. For example, if we see something suspicious and instead of a person going through manually this huge, huge sea container, we could spend 50 minutes and have a look what is actually inside.
00:09:53:17 - 00:10:05:19
Abigail Acton
Yeah. That's great. So it's like refining the scanning, taking it to the next level. Super. That's excellent. Thank you very much. Okay. And the pilot, was it well received? Were the people looking at it who worked for example, in customs, were they impressed by what they saw?
00:10:05:20 - 00:10:35:19
Siiri Salupere
Well, yes, it was impressive for many customs officers and that they do see the potential of it as a complementary technique. But the main challenge remains the long time for the scan. So what could be the potential way to overcome this? So maybe we could start making more muons so we wouldn't be dependent only on the natural flux of muons.
00:10:35:20 - 00:11:02:20
Siiri Salupere
So at sea level, the rate of natural muons is actually only one muon per square centimetre per minute. So there are not really that many. So in the next phase of the SilentBorder developments, we will be looking at a possibility to make more muons on-site using a laser plasma accelerator.
00:11:03:00 - 00:11:13:15
Abigail Acton
Right. Excellent. So if you could generate more, the whole process would go more quickly. Wonderful. Thank you. Thank you very much Siiri. That was beautifully explained. Does anyone have any observations or comments? Yes, Tarvo, please.
00:11:13:16 - 00:11:28:22
Tarvo Metspalu
So yeah, for listeners to understand the rate of muons flux, it is around – per one minute through our own bodies – as we have spoken here, around 600 muons have gone through you. So this might be an interesting fact.
00:11:29:00 - 00:11:50:07
Abigail Acton
Yeah. That is an interesting fact. And we're all still sitting here to tell the tale. So obviously they're also not harmful. So as Siiri was pointing out, thank you very much for that observation. The CosmoPort project set out to develop an innovative and highly effective solution for the next generation of scanner systems. So Tarvo, turning to you now, we're more familiar thanks to Siiri with regards to what muons are and what they can do.
00:11:50:09 - 00:11:55:10
Abigail Acton
So can you tell us what problem you were seeking to solve with this technology?
00:11:55:12 - 00:12:27:17
Tarvo Metspalu
Yeah. So whereas SilentBorder was aiming to see through sea containers, huge, huge containers, the CosmoPort is a sister project of SilentBorder aimed to have more of a mobile solution for smaller packages. The main thing we want to see through are postal roll cages in the logistics centres. So as you mentioned, those billions of parcels going through, then we would really love to see hundreds of parcels at the same time.
00:12:27:17 - 00:12:34:16
Tarvo Metspalu
And then having a more compact solution compared to SilentBorder is really kind of question to do that.
00:12:34:18 - 00:12:49:16
Abigail Acton
Right. Because my understanding is that the packages now are separated on a risk based analysis, which of course leaves a lot to human interpretation and so on. Is your project trying to help with the analysis of potential risk in any way?
00:12:49:18 - 00:13:14:04
Tarvo Metspalu
So what CosmoPort aims to do is that we want to find the most suspicious parcels in the postal roll cage to do further analysis on. So I know there are different methods out there, like digital noses or stuff like that. Chemical analysis and other methods, really good methods. But the main issue is that you have millions and millions of parcels.
00:13:14:04 - 00:13:29:21
Tarvo Metspalu
How do you decide on which of the parcel you do the chemical tests on? So the CosmoPort technology would help to bring out some more suspicious parcels to do the further analysis on, and then mark it down as illegal or not.
00:13:29:23 - 00:13:42:20
Abigail Acton
Right. Lovely. Because I understand that you are also involving AI and machine learning tools. So how does that work in practice? What is that an interpretation of the signal that you're getting from the scanner? Or how does that work?
00:13:42:23 - 00:14:09:13
Tarvo Metspalu
Using muography on parcels might seem like a straightforward task: you would see the density information, you would see the scattering information, and then you would just decide what is what. That is, at least in theory. Because in practice there are a lot of minute details. For instance, the hardware that we are using might have some anomalies.
00:14:09:19 - 00:14:42:24
Tarvo Metspalu
Also, there might be some flux anomalies. There are also some constraints which need to be taken into consideration. And moreover, if we do short measurements, then the images of objects might not pop up as you would be used to see in X-ray images. So that is why we have given the task over to AI, that we show the AI different types of mathematical calculations we have done on the muon scattering we see.
00:14:42:24 - 00:15:08:04
Tarvo Metspalu
And then we tell the AI that you should be able to find these kinds of objects, these kinds of materials, inside of the volume of interest. Do you see them and how well do you see them? And this we repeat several times through several iterations. And that is how we teach our ML algorithm to find different materials.
00:15:08:05 - 00:15:27:11
Abigail Acton
Wonderful. That's excellent. Thanks. Very clear. Thank you very much. The projects opening a door to a new system. I know this is very cutting edge and complex, and you're still working through various problems and hurdles that you're overcoming. So, you know, this is the start of a new idea. But how far did you develop the prototype?
00:15:27:12 - 00:15:32:05
Abigail Acton
What is the prototype that you developed in the project actually able to do at the moment?
00:15:32:07 - 00:16:18:18
Tarvo Metspalu
At the moment, the prototype is able to measure different parcels, and at the moment the prototype is able to differentiate specific material groups. Now not very, very specific and detailed chemical composition just yet. Because for instance, when you take carbohydrates like sugar, flour and, and stuff like that, then they are actually very similar in chemical composition. And that is where we still have a lot of development ongoing, but we are able to find different guns hidden between light parcels which come from AliExpress or Temu or something like that.
00:16:18:18 - 00:16:23:04
Tarvo Metspalu
So that is already quite the success we have had at the moment.
00:16:23:06 - 00:16:30:08
Abigail Acton
Perfect. Thank you. That's very well explained. Thanks very much. Does anyone have any observations or comments to make to Tarvo? So yes, Marco.
00:16:30:09 - 00:16:51:17
Marco Paleari
Thank you, Tarvo. Actually, I had a curiosity about how much defined can an image you extracted from muons can be. So how big is a single pixel? Can you detect a very small tiny object within a parcel or it has to be bigger?
00:16:51:19 - 00:17:24:06
Tarvo Metspalu
Thank you for the question. The main caveat of the current technology is that we are using natural muon flux. As Siiri mentioned, it would be very beneficial if we could generate additional muons. What this means is that we do not have enough information to get as clear of an image as we would get with an X-ray, but on the other hand, what we can do is that we can virtually define what is the resolution of the image.
00:17:24:10 - 00:18:04:00
Tarvo Metspalu
So if we would like to see one millimetre voxels so those are volumetric pixels, then we might get quite a detailed image over a few hours. So this means quite a long measurement time. But at the same time per this one millimetre pixel we wouldn't have enough information on what type of material is in that one millimetre. On the other hand, if we were to define in our mathematical models a big voxel like three centimetres, then the 3D image we would get would be more pixelated.
00:18:04:01 - 00:18:20:03
Tarvo Metspalu
So not so detailed, but at the same time, per one voxel we would have way more information about the material composition. So this is where we have to find the fine balance between resolution and material differentiation.
00:18:20:04 - 00:18:41:04
Abigail Acton
Perfect. Thank you very much. That's excellent. Yes, we can see that. Very well. Thank you. Marco, let's talk about your project. Dangerous and illicit goods, materials and substances flow across the EU in the mail, the iFLOWS project developed a new novel detection tool and also worked on improving the creation of intelligence to analyse risk and identify threats.
00:18:41:04 - 00:19:00:08
Abigail Acton
So we're looking at complementary systems here, the notion of muons complementing X-ray. Now we're also thinking about complementing both with aid to people with regards to what they should check and what they shouldn't check. So first of all, let's just start by why you were interested in this area of research in the first place. What attracted you to this this area, Marco?
00:19:00:10 - 00:19:38:13
Marco Paleari
Thank you, Abigail. To be honest, it's probably a mix of basically faith, me being a very curious person, buying stuff from everywhere and having a kid that became a junior engineer. It's a four year old. He's already playing with screwdrivers and bars and multimeters, stuff like this. And when you get into this kind of world, and I got in the project by chance actually when I got in engineering, and it resonated with me.
00:19:38:14 - 00:20:11:11
Marco Paleari
I mean, my kid finding a way to put his hands, his tiny hands on my stuff. And also, as you mentioned, we have a huge amount of parcels that are crossing Europe. You mentioned 4.6 billion parcels every day, which is about 140, 150 per minute, which it's impossible for our customs to actually analyse. So we have to find better ways to check and make sure that all parcels are secure and safe.
00:20:11:11 - 00:20:23:24
Marco Paleari
And of course, all of these has to be done without slowing down the industry, without destroying the packaging, possibly, and without invading one's privacy, which is also important.
00:20:24:03 - 00:20:37:22
Abigail Acton
Right. So it's a complicated puzzle to try and put together, and a balance to find. So iFLOWS wanted to achieve that, both through new technological approaches and helping officials make the right choices in the application of those approaches. So tell us more about that.
00:20:37:23 - 00:21:07:10
Marco Paleari
Yeah. In iFLOWS what we built was an integrated toolkit. The basic idea was to use AI as the brain to fuse information coming from different sources. And the first source would be the manifest. So the fingerprint of a parcel telling us where the parcel comes from, what is declared to be contained, how it was paid and so on.
00:21:07:12 - 00:21:45:23
Marco Paleari
But we also developed two new sensors to couple with the X-ray, which is legacy. The two sensors we developed were the Ultra Wide Band radar and the Terahertz scanner, the Ultra Wide Band radar basically gives us information about the radio frequency, active materials, which are included in a parcel so it can tell us if there are metal stripes, like in the cache, or if there are RFID chips like in credit cards, documents ID, public transportation cards, even in the anti-shoplifting labels nowadays.
00:21:45:23 - 00:22:09:01
Marco Paleari
And this information is already important. I think the real game changer is the Terahertz scanner. We devoted much effort on this. We have four companies developing this proof of concept. A Terahertz scanner - if I had to explain it - the closest way is to think about it as we think about the X-ray machines in the airport. Physically it's the same.
00:22:09:01 - 00:22:31:21
Marco Paleari
We have a conveyor belt and a tunnel. There's a sensor, there's a source on the top, a sensor on the bottom. It's not that's different from the muon scanner Tarvo and Siiri where discussing in that sense, but is smaller. But with the X-ray you will get basically a black and white picture about the density of what's hiden inside the parcel.
00:22:31:23 - 00:22:57:01
Marco Paleari
What we take is instead of doing a black and white photography, we are doing hyperspectral photography. How to get there? Let's think about it. X-ray is black and white. When we see things in colour, we do it because we have three receptors green, red and blue like in computer screens. What terahertz brings to the table is we have a thousand sensors.
00:22:57:03 - 00:23:24:12
Marco Paleari
That makes the image so much deep that we don't actually have a colour for an object, for a material, we have a spectral fingerprint. This spectral fingerprint basically is unique for every chemical compound. So for every pixel in our image, we will have the chemical composition of the pixel. Clearly this is not that simple. Again we have to use AI to process all of this information.
00:23:24:12 - 00:23:57:24
Marco Paleari
There are superposed material mixed materials in the same pixel. So all of these needs to be processed with AI. And indeed AI, as I was saying at the beginning, is the glue that we use across the whole project for analysing the shapes from the X-ray, from the Terahertz to detect the material or the content, to the Terahertz and Ultra Wide Band to fuse all of these information together, and at the end to give a single explainable threat score for the parcel.
00:23:58:01 - 00:23:58:12
Abigail Acton
Perfect.
00:23:58:12 - 00:24:10:24
Marco Paleari
That's excellent. Where explainable also is kind of important, because it means we're not only saying this parcel is dangerous, we're saying this parcel is dangerous because I think there is this material, or because I think there is this object.
00:24:11:00 - 00:24:28:15
Abigail Acton
That's fantastic. And then absolutely, because then the people looking at the information that's coming out will know which ones to open and which ones to look at so they can target their time and their energy, rather than just have this mass of stuff that they're just not quite sure which ones they should be looking at, apart from just using an X-ray, which yeah... very interesting.
00:24:28:15 - 00:24:34:18
Abigail Acton
Thank you. And super well explained to you. Does anyone have any comments for Marco or observations to make.
00:24:34:20 - 00:24:42:18
Siiri Salupere
Marco during your project, did you also develop a prototype of this instrument and what was the outcome of testing it?
00:24:42:19 - 00:25:20:24
Marco Paleari
Thank you. Siiri. Yes, we have the two prototypes, one for the Ultra Wide Band and one for the Terahertz. Actually, within the Terahertz we have three different prototypes one for the source, one for the detector, one for the whole thing. And we tested them in Linköping. We had a full scale experiment at the Swedish police, and we actually gave some custom officers from Greece the possibility to, not just interact, we really gave them parcels and the devices.
00:25:21:00 - 00:25:38:18
Marco Paleari
We instructed them on how to use them, and we had them analyse these parcels with the whole chain of tools, X-ray, Terahertz, Ultra Wide Bands and all of the software that is involved within. It was quite a nice experience.
00:25:38:20 - 00:25:41:05
Abigail Acton
Must made you very proud. So it went well. You were saying?
00:25:41:06 - 00:25:43:18
Marco Paleari
It went well. Better than expected.
00:25:43:19 - 00:25:52:17
Abigail Acton
Oh, that's a nice feeling. Good. Very good. It sounds like it's been a very intensive period of work, so you must be feeling quite proud to see that it actually works in practice.
00:25:52:19 - 00:26:19:22
Marco Paleari
Yes. In the end, it all worked out. Clearly we are talking about proof of concept. There's much more research that need to be performed. There is industrialisation and engineering that needs to be brought to bring the technology readiness level to a level in which this product can be actually applied in a real environment. But conceptually we demonstrated everything works.
00:26:19:24 - 00:26:31:11
Siiri Salupere
Marco, I have one more question to you: at this stage, at this proof of concept level, can you already give estimates? How long will it be the scanning times with your new instruments?
00:26:31:13 - 00:27:06:22
Marco Paleari
Yes, actually for the Ultra Wide Band it's very quick as radar it's about milliseconds. So it's just about how many we do want to take to have a more precise measurement. For the Terahertz, I'd say it's slower than X-ray, but kind of comparable. It all depends on how much power we can set in our source. We are using a quantum cascade laser, which is this innovative technology, to have a really focussed and powerful beam of light, Terahertz light.
00:27:06:24 - 00:27:17:22
Abigail Acton
Bravo. Bravo to all of you. That sounds excellent. Thank you so much for sharing your ideas. And also, you know, quite a technical concept that you've been exploring and you've all explained them so well. So thank you very much. Super.
00:27:17:24 - 00:27:24:10
Marco Paleari
Thank you, Abigail, for having us. It was nice talking to you and to Siiri and Tarvo.
00:27:24:12 - 00:27:27:24
Abigail Acton
Excellent. Thank you Marco. Thank you very much everybody. Bye bye.
00:27:28:01 - 00:27:30:12
Siiri Salupere
Thank you so much for inviting us.
00:27:30:12 - 00:27:32:00
Marco Paleari
Thank you. Goodbye.
00:27:32:03 - 00:27:35:05
Tarvo Metspalu
Yeah. Thank you for the opportunity. Really enjoyed it.
00:27:35:05 - 00:27:41:13
Abigail Acton
Very welcome. Goodbye, everybody.
00:27:41:15 - 00:28:00:12
Abigail Acton
If you've enjoyed this podcast, follow us on Spotify or Apple or wherever you get your podcasts. And check out the podcast homepage on the CORDIS website. Subscribe to make sure that the hottest research in EU-funded science isn't passing you by. And if you're enjoying listening, why not spread the word? We talked about how cognitive decline can be held at bay,
00:28:00:13 - 00:28:18:02
Abigail Acton
what deep time can tell us about future sea levels and tackling disinformation bots. Maybe you're involved in a project or would like to apply for funding. Take a look at the CORDIS website to see what others are doing in your domain. So come and check out the research that's revealing what makes our world tick. We're always happy to hear from you!
00:28:18:03 - 00:28:23:23
Abigail Acton
Drop us a line editorial@cordis.europa.eu.
Until next time.
This is an AI transcription.
00:00:15:16 - 00:00:48:20
Abigail Acton
Hello and welcome to this episode of CORDiscovery with me, Abigail Acton. Increasing amounts of dangerous and illegal substances pass through the EU, as the number of parcels being sent mount and shipping systems get more porous. According to the European Commission, the volume of small packages coming into the EU has doubled every year since 2022. In 2024 alone, 4.6 billion small packages entered the EU market, and this is not including the number of packages shipped within the EU itself.
00:00:49:00 - 00:01:21:12
Abigail Acton
How things are shipped is changing. These increasing volumes are now passing through postal agencies, private vehicles and small logistics companies. The analysis and detection capabilities in such locations are very low, so the process is a target for the smuggling of illicit goods. When it comes to bigger container shipments, ports are also a target. The European Ports Alliance gathers together operators, shipping associations, customs and law enforcement authorities to combat organised crime and drug trafficking by improving security imports.
00:01:21:12 - 00:01:34:11
Abigail Acton
But all advances in technology are very welcome. So how can the work being done by our three guests, all of whom have received support from the EU's research and innovation funding, have an impact?
00:01:34:13 - 00:01:51:09
Abigail Acton
Siiri Salupere is a research fellow in radiation protection at the University of Tartu in Estonia. Her work focuses on ionising radiation, looking at it as a potential hazard to human health, and also exploring the potential benefits of using cosmic rays for security applications. Welcome, Siiri.
00:01:51:11 - 00:01:52:23
Siiri Salupere
Hello. Great to be here.
00:01:53:00 - 00:02:08:13
Abigail Acton
Lovely to have you. Tarvo Metspalu is a product development engineer at GScan in Estonia, where he reviews hardware designs and machine learning data pipelines. He has been working on developing a more portable and efficient analysis and detection device for customs. Hi Tarvo!
00:02:08:15 - 00:02:10:14
Tarvo Metspalu
Hello. Thank you for inviting me.
00:02:10:15 - 00:02:35:08
Abigail Acton
Thank you for coming. Marco Paleari holds a PhD in effective computing from Telecom ParisTech, where he specialised in the intersection of AI and human emotions. He is currently a research and innovation project manager at the ‘AI data for digital defence’ unit at ENG, Italy. Marco describes himself as someone whose interests spanned just about everything, provided it contains a computer chip. Hi Marco.
00:02:35:12 - 00:02:37:11
Marco Paleari
Hello, Abigail. Glad to be here.
00:02:37:12 - 00:03:06:13
Abigail Acton
Very welcome. Siiri, I'm going to turn to you first, please. The SilentBorder project is developing a new high tech cosmic ray tomography, CRT for short, scanner that enables fast and safe screening, detection and identification of hazardous and illegal goods, contraband, and people hiding in containers. We're going to be talking a lot about muons today series. So can we just start by establishing what they are, what properties they have that make them useful as detectors?
00:03:06:15 - 00:03:45:22
Siiri Salupere
The muons are a part of cosmic radiation that is always surrounding us, and we can call muons the secondary cosmic radiation as they are produced in Earth's atmosphere when high energy particles from out of our planet bombard atmospheric particles, and then a chain of nuclear reactions occur and nuons are formed, they are actually elementary particles and quite similar to electrons, but much higher in mass.
00:03:45:22 - 00:03:54:01
Siiri Salupere
And that makes them distinct to electrons. They can penetrate more material than the electrons too.
00:03:54:02 - 00:03:55:08
Abigail Acton
Because they're bigger.
00:03:55:12 - 00:03:56:23
Siiri Salupere
Because they are bigger.
00:03:57:04 - 00:03:57:18
Abigail Acton
More massive.
00:03:57:18 - 00:04:30:24
Siiri Salupere
And they are more inert, so they can't be absorbed that quickly. And on the other hand, we could compare them to X-rays. And, in a more wider sense, we would call it electromagnetic radiation. Now, if we compare muons to electromagnetic magnetic radiation, the big change comes because they have a charge, which means that they scatter more than the X-rays would do.
00:04:31:00 - 00:04:38:11
Abigail Acton
So the fact that they have a charge, an electrical charge, means that when they impact on something, they scatter more?
00:04:38:14 - 00:05:05:19
Siiri Salupere
Yes. And now this scattering effect is dependent on the material. The higher the atomic mass of the material, the more muons are scattering, because there is an electromagnetic force between the positive charge of the nucleus of the matter and the negative charge of the muon.
00:05:05:20 - 00:05:25:10
Abigail Acton
So the muons are much more reactive when they come into contact with, when they hit, an object that's perhaps hidden or whatever, and the manner in which they react and how they scatter can give you information about the nature of the object that they're hitting. Is that correct?
00:05:25:11 - 00:05:43:10
Siiri Salupere
Exactly. And then if for X-rays, we could only distinguish material based on their density, then muons carry the information about the density of the material and also about the material composition.
00:05:43:10 - 00:05:52:09
Abigail Acton
Because they scatter in a certain way each time the composition changes. So if you can trace the patterns of the scattering, you have an idea of the composition of the material?
00:05:52:14 - 00:06:03:02
Siiri Salupere
Yes. And this can open up a window of being able to detect light materials which remain unseen for the common X-rays.
00:06:03:06 - 00:06:22:03
Abigail Acton
Okay, this is a fantastic explanation. Thank you so much. It's much clearer because, you know, we hear the word, but we don't necessarily know what it means or how it can be used practically. So now we've established a little bit about what muons are and why they're of interest to people who are trying to use them for the detection of materials in packages or in containers.
00:06:22:04 - 00:06:27:04
Abigail Acton
Can you then tell me a little bit about how your project harnessed this capacity?
00:06:27:06 - 00:07:02:12
Siiri Salupere
If we build a detector that has the possibility to detect the muon, and we have one detector plate on top of the object and another one underneath the object, then we are able to understand how the muon has been travelling through the matter, by the fact that we can understand how it has been scattering in the matter, and this enables us to detect what has been in the inside, the object that we have been scanning.
00:07:02:13 - 00:07:09:01
Abigail Acton
So this is great. So this is the technical capacity of muons. How did SilentBorder harness that?
00:07:09:03 - 00:07:47:02
Siiri Salupere
Yeah. So these objects inside the cargo we scan can possibly be something illegal, something that's not licensed. And why not even people. So this technology can help us to distinguish light elements. It can be something powdery. It can be a narcotic substance. And the current X-ray would not be able to see these differences because of the different properties of the X-ray and muons.
00:07:47:06 - 00:08:12:24
Siiri Salupere
And another fun fact about the muon technology is the fact that we are using a natural flux of radiation, meaning that we do not create any extra radiation dose which X-rays would produce. So even if there is a person hidden inside this cargo, they would not be getting any harmful effects from ionising radiation.
00:08:12:24 - 00:08:20:19
Abigail Acton
Right. So in the SilentBorder project, did you then construct a prototype to show how this could be used?
00:08:20:21 - 00:08:52:05
Siiri Salupere
Yes. The SilentBorder project built the prototype instrument of the scanner, and we also tried it out in real world trying to see the difference in light powdery materials like flour, starch, sugar, salt. And the technology really showed that it's possible to make the difference. All the building work was actually done by GScan, so you will hear more about GScan when Tarvo will be speaking.
00:08:52:05 - 00:08:56:13
Siiri Salupere
And our object to be scanned was a sea container.
00:08:56:13 - 00:09:07:02
Abigail Acton
So something very very big. So excellent. So how does this kind of work? You have a plate underneath and a plate above and you manipulate the container between the two like a giant sandwich? Or how does that go?
00:09:07:04 - 00:09:53:16
Siiri Salupere
Basically it is like a giant arch and the scanner is scanning what is underneath it, and the container is slowly passing through this scanner. Actually, this scanning takes quite a lot of time. It's 50 minutes, so we are not competing with the X-ray technology, but the SilentBorder technology could be a useful complementary technology to X-rays. For example, if we see something suspicious and instead of a person going through manually this huge, huge sea container, we could spend 50 minutes and have a look what is actually inside.
00:09:53:17 - 00:10:05:19
Abigail Acton
Yeah. That's great. So it's like refining the scanning, taking it to the next level. Super. That's excellent. Thank you very much. Okay. And the pilot, was it well received? Were the people looking at it who worked for example, in customs, were they impressed by what they saw?
00:10:05:20 - 00:10:35:19
Siiri Salupere
Well, yes, it was impressive for many customs officers and that they do see the potential of it as a complementary technique. But the main challenge remains the long time for the scan. So what could be the potential way to overcome this? So maybe we could start making more muons so we wouldn't be dependent only on the natural flux of muons.
00:10:35:20 - 00:11:02:20
Siiri Salupere
So at sea level, the rate of natural muons is actually only one muon per square centimetre per minute. So there are not really that many. So in the next phase of the SilentBorder developments, we will be looking at a possibility to make more muons on-site using a laser plasma accelerator.
00:11:03:00 - 00:11:13:15
Abigail Acton
Right. Excellent. So if you could generate more, the whole process would go more quickly. Wonderful. Thank you. Thank you very much Siiri. That was beautifully explained. Does anyone have any observations or comments? Yes, Tarvo, please.
00:11:13:16 - 00:11:28:22
Tarvo Metspalu
So yeah, for listeners to understand the rate of muons flux, it is around – per one minute through our own bodies – as we have spoken here, around 600 muons have gone through you. So this might be an interesting fact.
00:11:29:00 - 00:11:50:07
Abigail Acton
Yeah. That is an interesting fact. And we're all still sitting here to tell the tale. So obviously they're also not harmful. So as Siiri was pointing out, thank you very much for that observation. The CosmoPort project set out to develop an innovative and highly effective solution for the next generation of scanner systems. So Tarvo, turning to you now, we're more familiar thanks to Siiri with regards to what muons are and what they can do.
00:11:50:09 - 00:11:55:10
Abigail Acton
So can you tell us what problem you were seeking to solve with this technology?
00:11:55:12 - 00:12:27:17
Tarvo Metspalu
Yeah. So whereas SilentBorder was aiming to see through sea containers, huge, huge containers, the CosmoPort is a sister project of SilentBorder aimed to have more of a mobile solution for smaller packages. The main thing we want to see through are postal roll cages in the logistics centres. So as you mentioned, those billions of parcels going through, then we would really love to see hundreds of parcels at the same time.
00:12:27:17 - 00:12:34:16
Tarvo Metspalu
And then having a more compact solution compared to SilentBorder is really kind of question to do that.
00:12:34:18 - 00:12:49:16
Abigail Acton
Right. Because my understanding is that the packages now are separated on a risk based analysis, which of course leaves a lot to human interpretation and so on. Is your project trying to help with the analysis of potential risk in any way?
00:12:49:18 - 00:13:14:04
Tarvo Metspalu
So what CosmoPort aims to do is that we want to find the most suspicious parcels in the postal roll cage to do further analysis on. So I know there are different methods out there, like digital noses or stuff like that. Chemical analysis and other methods, really good methods. But the main issue is that you have millions and millions of parcels.
00:13:14:04 - 00:13:29:21
Tarvo Metspalu
How do you decide on which of the parcel you do the chemical tests on? So the CosmoPort technology would help to bring out some more suspicious parcels to do the further analysis on, and then mark it down as illegal or not.
00:13:29:23 - 00:13:42:20
Abigail Acton
Right. Lovely. Because I understand that you are also involving AI and machine learning tools. So how does that work in practice? What is that an interpretation of the signal that you're getting from the scanner? Or how does that work?
00:13:42:23 - 00:14:09:13
Tarvo Metspalu
Using muography on parcels might seem like a straightforward task: you would see the density information, you would see the scattering information, and then you would just decide what is what. That is, at least in theory. Because in practice there are a lot of minute details. For instance, the hardware that we are using might have some anomalies.
00:14:09:19 - 00:14:42:24
Tarvo Metspalu
Also, there might be some flux anomalies. There are also some constraints which need to be taken into consideration. And moreover, if we do short measurements, then the images of objects might not pop up as you would be used to see in X-ray images. So that is why we have given the task over to AI, that we show the AI different types of mathematical calculations we have done on the muon scattering we see.
00:14:42:24 - 00:15:08:04
Tarvo Metspalu
And then we tell the AI that you should be able to find these kinds of objects, these kinds of materials, inside of the volume of interest. Do you see them and how well do you see them? And this we repeat several times through several iterations. And that is how we teach our ML algorithm to find different materials.
00:15:08:05 - 00:15:27:11
Abigail Acton
Wonderful. That's excellent. Thanks. Very clear. Thank you very much. The projects opening a door to a new system. I know this is very cutting edge and complex, and you're still working through various problems and hurdles that you're overcoming. So, you know, this is the start of a new idea. But how far did you develop the prototype?
00:15:27:12 - 00:15:32:05
Abigail Acton
What is the prototype that you developed in the project actually able to do at the moment?
00:15:32:07 - 00:16:18:18
Tarvo Metspalu
At the moment, the prototype is able to measure different parcels, and at the moment the prototype is able to differentiate specific material groups. Now not very, very specific and detailed chemical composition just yet. Because for instance, when you take carbohydrates like sugar, flour and, and stuff like that, then they are actually very similar in chemical composition. And that is where we still have a lot of development ongoing, but we are able to find different guns hidden between light parcels which come from AliExpress or Temu or something like that.
00:16:18:18 - 00:16:23:04
Tarvo Metspalu
So that is already quite the success we have had at the moment.
00:16:23:06 - 00:16:30:08
Abigail Acton
Perfect. Thank you. That's very well explained. Thanks very much. Does anyone have any observations or comments to make to Tarvo? So yes, Marco.
00:16:30:09 - 00:16:51:17
Marco Paleari
Thank you, Tarvo. Actually, I had a curiosity about how much defined can an image you extracted from muons can be. So how big is a single pixel? Can you detect a very small tiny object within a parcel or it has to be bigger?
00:16:51:19 - 00:17:24:06
Tarvo Metspalu
Thank you for the question. The main caveat of the current technology is that we are using natural muon flux. As Siiri mentioned, it would be very beneficial if we could generate additional muons. What this means is that we do not have enough information to get as clear of an image as we would get with an X-ray, but on the other hand, what we can do is that we can virtually define what is the resolution of the image.
00:17:24:10 - 00:18:04:00
Tarvo Metspalu
So if we would like to see one millimetre voxels so those are volumetric pixels, then we might get quite a detailed image over a few hours. So this means quite a long measurement time. But at the same time per this one millimetre pixel we wouldn't have enough information on what type of material is in that one millimetre. On the other hand, if we were to define in our mathematical models a big voxel like three centimetres, then the 3D image we would get would be more pixelated.
00:18:04:01 - 00:18:20:03
Tarvo Metspalu
So not so detailed, but at the same time, per one voxel we would have way more information about the material composition. So this is where we have to find the fine balance between resolution and material differentiation.
00:18:20:04 - 00:18:41:04
Abigail Acton
Perfect. Thank you very much. That's excellent. Yes, we can see that. Very well. Thank you. Marco, let's talk about your project. Dangerous and illicit goods, materials and substances flow across the EU in the mail, the iFLOWS project developed a new novel detection tool and also worked on improving the creation of intelligence to analyse risk and identify threats.
00:18:41:04 - 00:19:00:08
Abigail Acton
So we're looking at complementary systems here, the notion of muons complementing X-ray. Now we're also thinking about complementing both with aid to people with regards to what they should check and what they shouldn't check. So first of all, let's just start by why you were interested in this area of research in the first place. What attracted you to this this area, Marco?
00:19:00:10 - 00:19:38:13
Marco Paleari
Thank you, Abigail. To be honest, it's probably a mix of basically faith, me being a very curious person, buying stuff from everywhere and having a kid that became a junior engineer. It's a four year old. He's already playing with screwdrivers and bars and multimeters, stuff like this. And when you get into this kind of world, and I got in the project by chance actually when I got in engineering, and it resonated with me.
00:19:38:14 - 00:20:11:11
Marco Paleari
I mean, my kid finding a way to put his hands, his tiny hands on my stuff. And also, as you mentioned, we have a huge amount of parcels that are crossing Europe. You mentioned 4.6 billion parcels every day, which is about 140, 150 per minute, which it's impossible for our customs to actually analyse. So we have to find better ways to check and make sure that all parcels are secure and safe.
00:20:11:11 - 00:20:23:24
Marco Paleari
And of course, all of these has to be done without slowing down the industry, without destroying the packaging, possibly, and without invading one's privacy, which is also important.
00:20:24:03 - 00:20:37:22
Abigail Acton
Right. So it's a complicated puzzle to try and put together, and a balance to find. So iFLOWS wanted to achieve that, both through new technological approaches and helping officials make the right choices in the application of those approaches. So tell us more about that.
00:20:37:23 - 00:21:07:10
Marco Paleari
Yeah. In iFLOWS what we built was an integrated toolkit. The basic idea was to use AI as the brain to fuse information coming from different sources. And the first source would be the manifest. So the fingerprint of a parcel telling us where the parcel comes from, what is declared to be contained, how it was paid and so on.
00:21:07:12 - 00:21:45:23
Marco Paleari
But we also developed two new sensors to couple with the X-ray, which is legacy. The two sensors we developed were the Ultra Wide Band radar and the Terahertz scanner, the Ultra Wide Band radar basically gives us information about the radio frequency, active materials, which are included in a parcel so it can tell us if there are metal stripes, like in the cache, or if there are RFID chips like in credit cards, documents ID, public transportation cards, even in the anti-shoplifting labels nowadays.
00:21:45:23 - 00:22:09:01
Marco Paleari
And this information is already important. I think the real game changer is the Terahertz scanner. We devoted much effort on this. We have four companies developing this proof of concept. A Terahertz scanner - if I had to explain it - the closest way is to think about it as we think about the X-ray machines in the airport. Physically it's the same.
00:22:09:01 - 00:22:31:21
Marco Paleari
We have a conveyor belt and a tunnel. There's a sensor, there's a source on the top, a sensor on the bottom. It's not that's different from the muon scanner Tarvo and Siiri where discussing in that sense, but is smaller. But with the X-ray you will get basically a black and white picture about the density of what's hiden inside the parcel.
00:22:31:23 - 00:22:57:01
Marco Paleari
What we take is instead of doing a black and white photography, we are doing hyperspectral photography. How to get there? Let's think about it. X-ray is black and white. When we see things in colour, we do it because we have three receptors green, red and blue like in computer screens. What terahertz brings to the table is we have a thousand sensors.
00:22:57:03 - 00:23:24:12
Marco Paleari
That makes the image so much deep that we don't actually have a colour for an object, for a material, we have a spectral fingerprint. This spectral fingerprint basically is unique for every chemical compound. So for every pixel in our image, we will have the chemical composition of the pixel. Clearly this is not that simple. Again we have to use AI to process all of this information.
00:23:24:12 - 00:23:57:24
Marco Paleari
There are superposed material mixed materials in the same pixel. So all of these needs to be processed with AI. And indeed AI, as I was saying at the beginning, is the glue that we use across the whole project for analysing the shapes from the X-ray, from the Terahertz to detect the material or the content, to the Terahertz and Ultra Wide Band to fuse all of these information together, and at the end to give a single explainable threat score for the parcel.
00:23:58:01 - 00:23:58:12
Abigail Acton
Perfect.
00:23:58:12 - 00:24:10:24
Marco Paleari
That's excellent. Where explainable also is kind of important, because it means we're not only saying this parcel is dangerous, we're saying this parcel is dangerous because I think there is this material, or because I think there is this object.
00:24:11:00 - 00:24:28:15
Abigail Acton
That's fantastic. And then absolutely, because then the people looking at the information that's coming out will know which ones to open and which ones to look at so they can target their time and their energy, rather than just have this mass of stuff that they're just not quite sure which ones they should be looking at, apart from just using an X-ray, which yeah... very interesting.
00:24:28:15 - 00:24:34:18
Abigail Acton
Thank you. And super well explained to you. Does anyone have any comments for Marco or observations to make.
00:24:34:20 - 00:24:42:18
Siiri Salupere
Marco during your project, did you also develop a prototype of this instrument and what was the outcome of testing it?
00:24:42:19 - 00:25:20:24
Marco Paleari
Thank you. Siiri. Yes, we have the two prototypes, one for the Ultra Wide Band and one for the Terahertz. Actually, within the Terahertz we have three different prototypes one for the source, one for the detector, one for the whole thing. And we tested them in Linköping. We had a full scale experiment at the Swedish police, and we actually gave some custom officers from Greece the possibility to, not just interact, we really gave them parcels and the devices.
00:25:21:00 - 00:25:38:18
Marco Paleari
We instructed them on how to use them, and we had them analyse these parcels with the whole chain of tools, X-ray, Terahertz, Ultra Wide Bands and all of the software that is involved within. It was quite a nice experience.
00:25:38:20 - 00:25:41:05
Abigail Acton
Must made you very proud. So it went well. You were saying?
00:25:41:06 - 00:25:43:18
Marco Paleari
It went well. Better than expected.
00:25:43:19 - 00:25:52:17
Abigail Acton
Oh, that's a nice feeling. Good. Very good. It sounds like it's been a very intensive period of work, so you must be feeling quite proud to see that it actually works in practice.
00:25:52:19 - 00:26:19:22
Marco Paleari
Yes. In the end, it all worked out. Clearly we are talking about proof of concept. There's much more research that need to be performed. There is industrialisation and engineering that needs to be brought to bring the technology readiness level to a level in which this product can be actually applied in a real environment. But conceptually we demonstrated everything works.
00:26:19:24 - 00:26:31:11
Siiri Salupere
Marco, I have one more question to you: at this stage, at this proof of concept level, can you already give estimates? How long will it be the scanning times with your new instruments?
00:26:31:13 - 00:27:06:22
Marco Paleari
Yes, actually for the Ultra Wide Band it's very quick as radar it's about milliseconds. So it's just about how many we do want to take to have a more precise measurement. For the Terahertz, I'd say it's slower than X-ray, but kind of comparable. It all depends on how much power we can set in our source. We are using a quantum cascade laser, which is this innovative technology, to have a really focussed and powerful beam of light, Terahertz light.
00:27:06:24 - 00:27:17:22
Abigail Acton
Bravo. Bravo to all of you. That sounds excellent. Thank you so much for sharing your ideas. And also, you know, quite a technical concept that you've been exploring and you've all explained them so well. So thank you very much. Super.
00:27:17:24 - 00:27:24:10
Marco Paleari
Thank you, Abigail, for having us. It was nice talking to you and to Siiri and Tarvo.
00:27:24:12 - 00:27:27:24
Abigail Acton
Excellent. Thank you Marco. Thank you very much everybody. Bye bye.
00:27:28:01 - 00:27:30:12
Siiri Salupere
Thank you so much for inviting us.
00:27:30:12 - 00:27:32:00
Marco Paleari
Thank you. Goodbye.
00:27:32:03 - 00:27:35:05
Tarvo Metspalu
Yeah. Thank you for the opportunity. Really enjoyed it.
00:27:35:05 - 00:27:41:13
Abigail Acton
Very welcome. Goodbye, everybody.
00:27:41:15 - 00:28:00:12
Abigail Acton
If you've enjoyed this podcast, follow us on Spotify or Apple or wherever you get your podcasts. And check out the podcast homepage on the CORDIS website. Subscribe to make sure that the hottest research in EU-funded science isn't passing you by. And if you're enjoying listening, why not spread the word? We talked about how cognitive decline can be held at bay,
00:28:00:13 - 00:28:18:02
Abigail Acton
what deep time can tell us about future sea levels and tackling disinformation bots. Maybe you're involved in a project or would like to apply for funding. Take a look at the CORDIS website to see what others are doing in your domain. So come and check out the research that's revealing what makes our world tick. We're always happy to hear from you!
00:28:18:03 - 00:28:23:23
Abigail Acton
Drop us a line editorial@cordis.europa.eu.
Until next time.
Spotting the difference between flour and cocaine – powerful new detectors are coming online
Dangerous and illegal substances pass through the EU every day due to porous shipping systems. And the number of packages is increasing. We are buying more and more items online and importing vast quantities of goods. How can we check what is being circulated, without damaging parcels, infringing on privacy or slowing down transactions? According to the European Commission, the volume of small packages coming into the EU has doubled every year since 2022. In 2024 alone, 4.6 billion small packages entered the EU market. And this is not including the number of packages shipped within the EU. How things are shipped is changing. These increasing volumes are now passing through postal agencies, private vehicles and small logistics companies. The analysis and detection capabilities in such locations are very low, so the process is a target for the smuggling of illicit goods. When it comes to bigger container shipments, ports are also a target: The European Ports Alliance gathers together operators, shipping associations, and customs and law enforcement authorities to combat organised crime and drug trafficking by improving security in ports. But all advances in technology are very welcome. So how can the work being done by our three guests, all of whom have received support from the EU’s research and innovation funding, have an impact? Siiri Salupere(opens in new window) is a research fellow in Radiation Protection(opens in new window) at the University of Tartu(opens in new window) in Estonia. Her work, explored through the SilentBorder project, focuses on ionising radiation and exploring the potential benefits of using cosmic rays for security applications. Tarvo Metspalu(opens in new window) is a product development engineer at GScan(opens in new window) in Estonia where he reviews hardware designs and machine learning data pipelines. Through his CosmoPort project, he has been working on developing a more portable and efficient analysis and detection device for customs. Marco Paleari(opens in new window) holds a PhD in Affective Computing from TelecomParisTech, where he specialised in the intersection of AI and human emotions. He is currently a research and innovation project manager at the ‘AI data for digital defence’ unit at ENG(opens in new window), Italy. Paleari coordinated the iFLOWS project.
Happy to hear from you!
If you have any feedback, we’re always happy to hear from you! Send us any comments, questions or suggestions to: editorial@cordis.europa.eu.
Countries
Austria, Estonia