This episode explores SiLC Technologies' Eyeonic Vista, an ultra-long-range 4D LiDAR vision system built on silicon photonics that provides wide-area coverage beyond a kilometer with per-pixel distance, velocity, and polarization measurements.

Host Jim Tatum speaks with Arlon Martin, SiLC Technologies Vice President of Market Development, about the system's features, including dynamic region-of-interest scaling, object detection and identification capabilities, integration with radar and cameras, performance in fog and dust, and how the system supports high-end security and layered sensing solutions.

Related: SiLC's 4D Vision System Brings Kilometer-Range Drone Tracking into Focus

Related: Drone Detection System Uses AI to Fight Prison Contraband Smuggling

Visions: A Machine Vision and Automation Solutions Podcast, is the podcast for engineers, designers, integrators, and end users who want to keep an informed eye on the imaging and machine vision industry. Every second and fourth Tuesday we will explore the latest in imaging trends, developments and solutions. Here you will find interesting, useful insights and observations from expert interviews, solo episodes, even the occasional panel discussion, all of which aim to expand your knowledge on imaging and machine vision. 

Transcript

Well, hello and welcome to Visions, A Machine Vision and Automation Solutions podcast. I'm your host, Jim Tatum, senior editor of Vision Systems design and Visions is an Endeavor Business Media production from your friends at Vision Systems Design. Here you'll find the latest on everything from end user machine vision solutions to trends, developments, and perspectives on all things machine vision and imaging. Whether you've been working in the industry for a while or you're just starting to take a closer look at it, this podcast is designed to grow your knowledge and bring greater focus to your understanding of the imaging and machine vision industry. And now on to our show.

Well, hi everybody, and welcome to visions. Today we're going to take a closer look at a really interesting development in long range sensing, the Eyeonic Vista system. This is an ultra-long range 4D LiDAR platform that's doing something pretty unique. It's combining wide-area coverage with dynamic region of interest scaling plus per pixel velocity and polarization data, all at distances beyond a kilometer. The system, as you may guess, is essentially designed for high end security. But as you'll no doubt find out shortly, there's quite a bit more going on here. Our guest today is Arlen Martin, who is vice president of market development at SiLC Technologies, a silicon photonics and machine vision innovator based out of Monrovia, California, that specializes in this technology and has developed the Ionic Vista system. Welcome, Arlen, and thanks very much for spending a few minutes with us today. Um, if you would, for starters, just please tell us a little about first about yourself and then a little bit about the system and what it does.

Yeah. This is Arlen Martin and I have been working with the silk team through several different generations of products for telecom and dad come, but it's silk. We have really focused on vision systems using silicon photonics. And the silicon photonics, uh, gives us a whole lot of capability that just isn't available in other systems. And so today, the focus is really on one of our new products that we just announced, uh, and it's called the Ionic Vista Vision System. So we're delighted to be here to, uh, to talk a little bit about it and about some of its capabilities.

Okay. First and foremost, what does it do?

So typically you would put this on a pan tilt and you would look at, uh, images for either border security where you want to see maybe people that are traveling. Uh, trafficking or unknown people coming across the border. You might use it for port security where you want to look at the marine traffic coming in and out of a port to see if there are vessels that you don't recognize that you would want to have a closer look at. Uh, you would use it for looking at drones. So this has a very good capability and very good resolution. And some of the features of it make it very, very useful for tracking and identifying drones for, uh, security. 

Okay. Okay. Um, somewhere in there, I read that you guys are mentioning dynamic ROI. Scaling is a key differentiator. Can you walk us through what that looks like in a real deployment? For example, tracking a small drone while still monitoring the whole area. 

The field of view for our vision system is, uh, thirty by sixty by thirty degrees, which is a pretty wide field of view at a kilometer, which is, you know, just using that as a reference point. That means a field of view is like a kilometer plus wide and about more, a little bit more than a half a kilometer high. So that's a very large field of view. If you're looking, you know, over a stadium or an airport or something, when you see something which might be a drone, or it might be a bird that you want to have a closer look at, then you scale it down to have a very short field of view to get more of your visibility, just like a pair of binoculars or another camera, you scale it down and you can get to a range of, of, you know, maybe something like twenty meters by forty meters. So you've gone from, you know, a kilometer in one direction to a half a kilometer in the other direction down to something which is like a twenty by forty. And now you can really zero in on, on what it is that you want to look at and get a very high resolution view, a horizontal and a vertical resolution at that is like on the order of fifteen centimeters. So, you know, we can look at objects, we can see objects that are less than a half a foot in terms of size. And so that gives us a very good a very high resolution at at those ranges. One of the things that's unique about our vision system is that while we're doing scanning and we're sending those pixels back to our camera, to our vision system, we can measure the distance with high accuracy of every pixel. That means we know exactly how far away that object is and the speed of it. Our accuracy of measurement at a kilometer is like ten centimeters. So we we know exactly what radar can't do this. A camera can't do this. A camera doesn't know how far away objects are. Even with a rangefinder. They just give you a general idea. But the reason you want to know exactly how far it is, is that if you want to use something to intercept that drone or a high, let's say you pair our system with a high energy laser. The laser needs to know exactly how far away that object is in order to be useful. And so what our system would do in that case would be provide you the distance very accurately. The other thing that we provide is the velocity, the speed of it. So we can not only tell you how far it is, but we can tell you which direction it's going and how fast. With every pixel, we provide that velocity information. And that's why this is a really very powerful, very unique vision system that didn't exist before. 

Wow. Um, how fast can that ROI adapt frame to frame and what trade-offs, if any, are happening behind the scenes? 

I don't have the the number for exactly how fast, but it's, it's very fast. So you can see something you can focus on it. You could, you know, uh, maybe you see three or four objects if it's a swarm. And in that case, since you know the velocity of the different objects, you might then prioritize them into one, two, or three. The one that's coming at you fast and the one that's. And uh, uh, maybe your highest priority. And then you pass that off to some other system to decide whether to use an RF jammer to send it home or send it to the ground or a high energy laser to destroy it. So there's, there's lots of things. So after you've picked your first target and decided what to do with it, then you move to your next one and to your next one in an order. 
So these decisions are being made by a human operator. Is it AI based in the early days, which is where a lot of these things are, they're still, in some cases, uh, being made by a human operator. But the big difference here is our capability to distinguish between things that might be friendly like a bird. Right? And, and, um, we do that, uh, these are things that radar and cameras have a hard time doing at long distances is to figure out whether it's a bird or a drone. We do that because we, uh, detect both polarization modes of light. And not to get too technical here, but the laser will send out light in, um, one mode and one polarization. And when it strikes an object that is kind of smooth or more shiny or more drone like, it'll reflect most of that light in the same polarization mode that it sent. But when that light hits a bird with its feathers and with the movement of the bird, that polarization gets scrambled and when it comes back, it could be instead of tea, it could be tea and TM in some combination. And by looking at the amount of light which is converted from tea to TM, that tells us whether it is a smooth object and looks like a drone, or whether it has been scrambled and is a bird, which we really don't want to take action on.

You say this this system really has not really been created before. So are there things that are being done now? For example, you know, you're talking about is it a bird or is it a drone? Were they having extreme difficulty doing that before? 

I mean, lots of false alarms. And we want to mechanize this whole thing. And in order to mechanize, it means that the probability has to be like, you know, ninety nine percent Right? Or more. And so to be able to automate this whole process, we need to do it very carefully. One of the problems with with radar is that it will lose track of a drone if it hovers or tries to get sneaky and goes slow. And ours will track it, uh, even if it stopped. We know where it is and we can watch it. We actually see, in many cases, the propellers rotating of the drone because we do velocity detection. Uh, we know where it is. And if it's stopped, we know where it is. Another problem with radar is, is if the drone drops down below the horizon, or there's a lot of background close to the ground, the ground will give a lot of reflections and create a lot of noise, and we'll lose track of it. Our leader can track a drone, you know, Feet off the ground or wherever. It doesn't matter where it is in terms of, you know, close to the ground stops or whatever. So those are cases where radar does a kind of a poor job. And then we operate, uh, with invisible light and we operate day or night. So it makes no difference for us. We can see it just as well at night as in the daytime because we're just using, uh, fifteen fifty nanometer light. Uh, that's invisible to us. 

Okay. Without giving away trade secrets or anything, can you kind of give me an idea of the components in the system? I mean, it's a forty forty radar, but it's also being integrated with camera systems or not operating separately. 
You know, those those would typically be separate. They're not integrated with the camera. The camera can work well in certain lighting Conditions, but there are certain other lighting conditions where cameras fail miserably, and we don't see because we're looking only at our own light. And it's a coherent detection glare doesn't matter to us. It you know, you could shine sun straight into our product. It doesn't matter. Looking straight into the sun. Uh, you can have multiple leaders in the same field and they won't interfere with each other. And so we have a very noise free vision system picture of whatever it is we're looking at. And that's very, very useful. Uh, another thing that was mentioned stand out capability is per pixel velocity and micro Doppler. And so for an engineer used to radar plus stacks, what changes when that data is embedded directly into a vision system. So with radar, we kind of get velocity too. And in resolution two. But the wavelength of light at fifteen fifty is about a thousand times smaller than the radar wavelength. Of course, it depends on which radar, but let's use one thousand for times, and that gives us a thousand times better resolution of the speed of the velocity. And so that allows us to see things like even the motion of the propellers on a drone, if it's within range, and it allows us to look at the velocity signature in addition to the polarization mode, to say, oh, this thing is flapping its wings. That's not a drone. And so it's another dimension to the AI portion of what it is. Are we seeing and is it friendly, something that we're okay with? Or is it something that could be dangerous? Okay, so for integrators building layered systems, where does this fit? So it can be used standalone, but in many cases, it would be used in combination with radar or with a camera. The radar or camera would say, you know, we're seeing something out there. We need to have a look at it. And so the radar would maybe pass off, you know, roughly that wide field of view as to where to point the lidar. And then the LiDAR would then look and drill on it. And we'd use a dynamic field of view to scope in and to gather the information that we want and then send that on back to the system. A full system stack would then, you know, decide what action to take and really make the decisions of what to do about the threat. 

Um, well, operating at one thousand five hundred and fifty NM, I can't pronounce it. NM nanometer. Yeah. Multi kilometer scenarios. How does the system hold up in things like fog and dust? Bright sunlight. You don't seem to have a problem at all with any of that. 

Yeah. So we looked at this with dust. We've done dust tests. We've done fog tests. And all of them do degrade. But we have a whole lot better visibility than we do with the human eye or with other with camera sensors in particular. Um, and that's partially because we can do things like range gating because we know the distance of every pixel. So that means that with a little AI, we can remove a lot of the noise from fog or, or dust or whatever, and we can focus in on the object which is moving and which is where we know that the the distance is it's still up in the sense that we still have a degraded signal, so it won't look as clean if it's a cloud. But definitely we can do things like see a drone through a cloud, um, follow it, etc. realistically, how somebody looking at this thing, how much degradation might they expect to actually have a situation? Well, this is where it comes into the degree of smoke and the degree of, uh, just like all of us, uh, you know, is it a hazy day or is it, uh, full blown smoke or is it fog? various things. So we need to do more in terms of getting more data. But you know, anecdotally, if you know it, it could reduce the range. So instead of, you know, maybe we're seeing things at a kilometer and a half on a clear day and then a foggy day. Instead of that, it's like at a kilometer. So it can have some effects like that. One of the other points that I didn't really mention about the Vista is that we have eight simultaneous beams, so we're sending out like eight channels at a time. And part of the purpose of that is to give us more pixels coming back. So if we can send out eight times as many beams of light, we can get up, you know, on an object or something, we can get eight times as many photons, you know, coming back pixels coming back. And that's important for a lot of these corner cases where the visibility might be impacted by, you know, something with eight channels like that and the high res four D output. 
I'm an engineer. How do I think about what belongs in the edge versus upstream. So we provide a point cloud. And part of our physical AI is to not destroy the data, which means that we want to provide all of the data rather than having it filtered or screened in some kind of way in the point cloud. And then when we run, you know, AI or other programs on the point cloud to do the decision making and do the whatever, we have more features at our control. More features with every pixel to look at, we can look at the velocity, we can look at the distance and we can look at the size of it, the x, y. So we have a lot more data that we can use in terms of processing the image.

Well, that's a wrap for this episode of visions produced by Endeavor Business Media, a division of endeavor B2B. Thanks very much for tuning in. If you enjoyed today's show, be sure to subscribe to the podcast and share this episode with a colleague who would find it helpful. Until our next episode, you can find us at vision dash systems dot com or on LinkedIn, Facebook, or for more insights, updates, and breaking news to keep you in the know. Thanks for tuning in. Until next time, stay focused on your visions.