Member Info for TheLongestShot

I can report that no auto plants in Aberdeen have the UAW on strike!

As I suspected. Gentex use a vapour sensor to confirm intoxication

Tom Toma 2nd

Product Management at Qualcomm

1d Edited

Follow

Check out my latest patent together with Thomas Herbert that highlights the potential to use the in-cabin sensing together with sensors on the outside or looking to the environment of the vehicle for adapting control and warnings from the Automated Driver Assist System using a dynamic attention area. #adas

https://www.linkedin.com/posts/tom-toma-29ba20a_adas-activity-7108467205549281280-Cmzi?utm_source=share&utm_medium=member_android

Try that again.

Not keeping all it's eggs in one basket, this one from Magna isn't in the mirror, but on the dash and integrated with someone like Mobileye. Since they mention Mobileye, this isn't just a placeholder to stop others going this route

Not keeping all it's eggs in one basket, this one from Magna has the mirror on the dash and integrated with someone like Mobileye. Since they mention Mobileye, this isn't just a placeholder to stop others going this route.

Pretty indecipherable, but they have added some parameters to measure driver state. Nothing actually about how the data is captured so not a worry for us

I think someone has a shorting agenda. I won't say who of course. No one can be spoon fed this many facts and still find arguments. It is blobby time. I have a strange feeling we have met this type before.

LittleFella, I think you are confusing DMS - Driver monitoring with actually controlling the steering and braking which Seeing Machines DO NOT do.

When we are excited to see sales of Blue Cruise or Supercruise, it is because Seeing Machines is checking the driver is paying attention, unlike in a Tesla where they don't need to actively monitor.

I far prefer the European approach of installing DMS across the board - helping to get everyone home safe.

Hanks JC, another OEM trying to differentiate themselves by adding ingredients to a standard sauce.

Off the top of my head, salience is how long the eye stays on or tracks an object. So the vehicle is looking forwards identifying objects in the road area that are relevant to driving and is checking that the driver being monitored is adjusting their gaze actively to check relative vehicles. So they know the driver is still actively alert and not just staring blankly in the correct direction

The camera and illumination will cover a broad area. Initial Driver Monitoring Systems had to cover just the driver area but the accommodated tiny grannies to roof scraping, lanky loons and the actively scanned width would be just as accommodating.

With OMS the driver and passenger area is covered easily so head and body movements are not a problem.

The moveable mirror was a bigger challenge, but long since solved.

The simple option is the 5-10° shift when you use the manual dip. This small offset moves the reflected headlights from you eyes down to your neck or chest.

The camera hasn't moved much and quickly finds your face and eyes and carries on.

However, the DMS also needs to know where YOU are looking, so first it needs to know where IT is looking.

In a car, the DMS is either in a fixed location or in the movable mirror. In a fixed location it always knows where it is facing and has set parameters for where the driver can be. Not forward of the steering wheel, or further back than the chair will adjust. It knows the range of human face sizes, so can estimate the distance to the eyes from the camera. Some trigonometry gives the position of the eya relative to the camera and also to the 3d model of the car. So now with a gaze vector you can calculate if the driver is looking at instruments, mirrors, windscreen, inside the car etc.

So the mirror option also has to identify some fixed features in the car like the B-pillars, or handles above passenger doors or light housings in the roof liner. Now it starts the trig calcs to identify its Point of View, then it can place the driver in the vehicle 3d model.

The camera and illumination will cover a broad area. Initial Driver Monitoring Systems had to cover just the driver area but the accommodated tiny grannies to roof scraping, lanky loons and the actively scanned width would be just as accommodating.

With OMS the driver and passenger area is covered easily so head and body movements are not a problem.

The moveable mirror was a bigger challenge, but long since solved.

The simple option is the 5-10° shift when you use the manual dip. This small offset moves the reflected headlights from you eyes down to your neck or chest.

The camera hasn't moved much and quickly finds your face and eyes and carries on.

However, the DMS also needs to know where YOU are looking, so first it needs to know where IT is looking.

In a car, the DMS is either in a fixed location or in the movable mirror. In a fixed location it always knows where it is facing and has set parameters for where the driver can be. Not forward of the steering wheel, or further back than the chair will adjust. It knows the range of human face sizes, so can estimate the distance to the eyes from the camera. Some trigonometry gives the position of the eya relative to the camera and also to the 3d model of the car. So now with a gaze vector you can calculate if the driver is looking at instruments, mirrors, windscreen, inside the car etc.

So the mirror option also has to identify some fixed features in the car like the B-pillars, or handles above passenger doors or light housings in the roof liner. Now it starts the trig calcs to identify its Point of View, then it can place the driver in the vehicle 3d model.

Hi LittleFella,

Your question about "How would the rear-view mirror based cameras cope with being dazzled by full beam headlights from the vehicle behind?" is a good reminder that often key parts of the technology are not understood by the wider public.

These are cameras that are using near infrared and they also have LED or 'vertical cavity lasers' operating at 940 nm. This particular wave length of illumination is important. First it is invisible to humans (earlier generations of infrared LEDs used to also have a faint visible glow). Secondly, the sun generates lots of infra-red light but at 940 nanometers it is filtered out by the atmosphere, so hardly any reaches the surface. So the camera won't be dazzled by the sun in infrared.

The third part is for your question and headlights, whether LED or traditional glowing light bulbs can't dazzle in IR if they don't emit that wavelength.

A further feature is there are filters that block unwanted ranges of the light spectrum. Where there is a combined RGB-IR camera individual pixels will 'see' only red, green, blue or IR.

The camera outputs either a colour 'visible' image or the 'invisible' image created from the IR illumination only.

How to detect if the human is paying attention - the camera picks things in the road ahead and then waits to see if your eye follows them, if so then you are alert enough to drive or monitor the vehicle

Thanks JC,

Hope this means that SEE have confidentially added rPPG to their list of optional features. It detects slight differences in skin reflectivity caused by pulse.

That means that the frame rate must be faster than the pulse (at least periodically) to detect reliably.

Warning - may not work for clowns in full pancake face makeup!

As I see it, they are taking on engineers, testers, accounts staff, lawyers etc. The first are required not for Guardian Gen 3 which will now be getting initial production batches tested in the wild with partners.

Instead those new teams are busy working on large numbers of NRE projects for existing and new customers. These will be subject to NDA and not material to the business on their own so have not been announced. Only when the NRE is delivered and integrated with the Tier 1s physical set up will the contracts that we are waiting for get announced. V & V maybe required before we get there, so it is good that we are tied in to Devant.

So Paul is issuing all the recruitment videos he can, because a steady supply of Engineers is required to keep the company growing. His Cheshire cat grin and swagger, tells you the success of the policy. He can't spell t out himself. But remember we are on our own on the final straight. We don't have each of the contracts yet. But we will win 50% and then some

Thinking back to yesterday's video and today's they both show the professionalism of the company, it doesn't have growing pains of a small company that makes mistakes.

They have hired with a big company mentality. They are scaling up to be the multibillion dollar company. Which means lots of boring parts. But to me that is de-risked money

It probably doesn't help the sp that we don't see or experience GM since they sold Vauxhall and Opel. It will help when we get more exposure to other handsfree systems from Stelantis and VW, not just the premium German OEMs

there are a few things that will make aviation easier than automotive. no 16 year old or 80 year old. pilots dress and hair is pretty conservative as are their glasses and hats. they have to pass regular medicals and will have two working eyes without a s*****. min and max height rules constrain some more excesses. if a pilot looks drunk or asleep while sober and awake due to other factors they wouldn't get the job in the first place.

in the cabin they might have lightning but no tunnels, trees or rear windows to make lighting difficult. there may be vibration, but no 'belgian pave'

Not so much DMS, but uses multiple cameras to generate a picture of the driver's face from a view point that is directly face on

JC, this is a good example of how to beat a patent troll. Actually build something and record specific steps to find a unique approach. Veoneer are the good guys, they want to save lives.

It starts off simply with using a vertical plane to represent a "photo" of the scene in front and plots where the gaze intersects with what is seen.

But then the claims show how this viewpoint needs to keep changing based on road curvature and the number of road lanes etc