HOME  |  NEWS  |  BLOGS  |  MESSAGES  |  FEATURES  |  VIDEOS  |  WEBINARS  |  INDUSTRIES  |  FOCUS ON FUNDAMENTALS
  |  REGISTER  |  LOGIN  |  HELP
Comments
You must login to participate in this chat. Please login.

hello all from Edmonton, Alberta.

Iron

Good course topic to help force a review of the issues and state of the art and future of the field ... thanks for archiving courses generally ...

This five day course could have been a [the] capstone seminar.  A lot of excellent topics and well versed presenters helped tie the course / seminar together.  However, as a whole I would ask the Design News Content Director and Embedded Vision Alliance folks to review the approach and content -- starting with the intended audience.  The depth of coverage, the sequence, starting with the introduction, a short history of the vision field (just a few slides would help), the tie to current state of the art, the migration path and timeline, and details of applications would make this a wonderful course to review.  Otherwise just a lot of teasing, hinting, and alluding to concepts that really need better coverage.

Thanks Simon for a great talk and everyone for a interesting course.

Iron

For those of you interested in learning more about embedded vision, I recommend the website of the Embedded Vision Alliance, www.embedded-vision.com, which contains extensive free educational materials. 

For those who want to do some easy and fun hands-on experiments with embedded vision first-hand, try the BDTI OpenCV Executable Demo Package (for Windows), available at www.embedded-vision.com/platinum-members/bdti/embedded-vision-training/downloads/pages/introduction-computer-vision-using-op

And for those who want to start developing their own vision algorithms and applications using OpenCV, the  BDTI Quick-Start OpenCV Kit (which runs under the VMware player on Windows, Mac, or Linux) makes it easy to get started: www.embedded-vision.com/platinum-members/bdti/embedded-vision-training/downloads/pages/OpenCVVMWareImage

Blogger

Thanks Simon's presentation!

Iron

Thank you Simon and all participants.

Thank you from me too. Bye.

Iron

thank you guys again, goodbye

Iron

Thank you for the answers. Goodbye everyone.

Iron

Thank-you everyone for all the questions and for listening today.  I look forward to perhaps meeting some of you at the Embedded Vision Summit in Boston on 25 April.  I am signing off.  Have a great weekend.

Blogger

Simon, thank for the answers, I will look into AVB

Iron

QUESTION:  Regarding collaboration with universities: Are embedded vision projects very suitable for student projects/graduations? In the previously mentioned design flow, I can imagine that the high modularization is well suited for collaboration.
ADiewi

ANS:   Yes.  We actually have a face recognition and gesture recognition research project in place with students at Ottawa University and would like to support expansion of this type of research work at other universities.   Colloboration works well where the university is involved primarily on the algorithm analysis and sometimes in helping set up initial image test base and then the optimization and productization is managed primarily by the partner company. 

Blogger

Hmm, I wonder if backing up over some very broken up or noisy concrete might fool the system; OR if a chalk drawing of a bike complete with shadow or human had been drawn on the concrete would fool it.  Probably ok for the driver to get occasional false positives.  I don't suppose the system can tell the difference between a flat image and a real 3D object of same?  The latter would change differently while getting closer to it.

Iron

QUESTION:  thanks for the clarification Simon, was raw Bayer data or processed RGB data used from the cmos sensor?
noor811

ANS:   The sensors are commonly configured to provide YUV colour format image frames to the image cognition processor and so this was the raw form of the image data processed.

Blogger

Historically, all inventions mostly come by accident of some sort (problems), so problem has to be worked out to become a solution. Therefore, each error it just scaled down to microlevel  problem in SW/HW and we really do not know how to treat that event as  "deadly error" or "potencial invention on microlevel".

QUESTION:  How is your hardware and software reliability in the field?
mrybczyn

ANS:  Good question.  The chip used is automotive qualified and highly reliable and needs to meet low DPM. Freescale is delivering this type of chip, called SCP2201.   The sw platform and library is also highly reliable and production proven.

We do not know the customer's physical module design hardware  reliability, but we have seen at least in the prototype and pre-production designs by different customers a range in reliability. 

 

Blogger

QUESTION: Simon, would an FPGA have been a better alternative, maybe not cheaper?
noor811

ANS:   Not likely in this case.  Power needed to be under 500mW at 85C for after-market and for before market under 105C so the power at room temp really needs to be < 350mW.  The chip used also needs to pump out analog NTSC video.  Furthermore, some customers are implementing their own image and vision processing algorithms while others are looking for a tool kit or almost turn key solution to start with.  The chip footprint including system memory needed to fit into <100mm2 area.

Blogger

thanks for the clarification Simon, was raw Bayer data or processed RGB data used from the cmos sensor?

Iron

Question for wonohkim:  Did you really mean lightning or lighting?  Is your system weather related?  Or does it need to handle the rare case of lightning strikes, oversaturating the video, anyway?

"In my case the key problem of machine vision was lightning"

Iron

QUESTION:  Hi Simon, could you please tell me about feature detection, are you able to diferentiate humans, animals, objects, etc?
fdarocha

ANS:  No.  There is no classification being done in this case.

Blogger

QUESTION:  Simon, why was an analog camera used rather than digital/CMOS?
noor811

ANS:  Actually a digital CMOS sensor was used, but the de-warped raw digital video with overlay is sent to the display in the car by twisted pair as NTSC analog. This is the most common and cheapest method.  We have seen new modules designed that can send araw digital video over NTSC and for before market some automotive OEMs are moving to use Ethernet AVB to connect the cameras to the display.

Blogger

@wahaufler: Very true. We need to remind ourselves what software engineers tell in every lecture/presentation. If an error is not detected in the step where it emerges, the cost to remove it will increase by 10 through every additional step in development. This not very different for us.

Iron

QUESTION: What criteria is for golden image test database? Is it specific to the field?
Anatoliy1086

ANS:  None and customers did not specify any except there is a guideline for testing ultra-sonic sensors.  We tried to build a robust test data base by thinking of all the possible environmental and use case conditions (indoor, outdoor, multiple objects, people, cars etc...) and create a test video clip that addresses the condition.

Blogger

@SCMorris:

Regarding collaboration with universities: Are embedded vision projects very suitable for student projects/graduations? In the previously mentioned design flow, I can imagine that the high modularization is well suited for collaboration.

Iron

QUESTION: Is there a Pedestrian detection library? Does it meet safety guidelines?
sarimarcosa

ANS:  Not for this type of implementation.  The customers did not want to detect only pedestrians so classification of objects as pedestrians did not need to be done. The customers wanted to detect and warn the driver that the car is going to hit something regardless.    There are European and US NHTSA safety guidelines that we are aware of, but I am not aware of any that currently specify that the rear camera needs to detect objects, determine distance and warn the driver.  This requirement is being driven mostly by after-market customers and automotive OEMs who want to provide this safety feature as a differentiator.

 

Blogger

Us engineers, by nature, tend to want to dive into a problem.  It takes discipline to first plan and design.  It takes more discipline, I think, to consider the problem from its very basics, first.

Iron

Testing , testing .... and testing

@wahaufler:

Your final statement is very true. The great engineers see the main aspects in the front. But even they miss many aspects. The difficulty is to gather the critical points that are able to endanger a project, which takes a LOT of experince to even recognize them as critical. Sometimes, it's even impossible to do so.

 

Iron

my first goal is to do a laser pointer detection algorithm

Iron

In my case the key problem of machine vision was lightning

Iron

The small Spartan-6 LX45 FPGA Atlys Dev board which I'm running a reference design that interfaces with Aptina CMOS image sensor, uses DDR2 for frame buffer, and displays on HDMI monitor, gets quite hot, needs a heatsink..

Iron

Can you describe your quality & reliability processes? For hardware specifically

Iron

Assumption has to be the #1 enemy in development of machine vision. ;)

QUESTION:  what was the development effort (staff hours)?
caa028

ANS:  Good question and hard to sum up as some of the work ended up being applied to provide other products such as Image Processing Algorithm Tool Kit or PC based Look-Up Table (LUT) Generation Tool.  CogniVue is only a 30 person company with a team of 24 R&D so it consumed a huge portion of that team' time for about 18 months.  We also colloborated with the University off Ottawa Vision lab so some grad students assisted during this period.

Blogger

@ADiewi, thanks, my questions are open for anyone to answer

Iron

@noor811:

This is most likely a question of size, power supply and heat dissipation. FPGAs are as far as I know overdimensioned for deeply embedded systems.

(I hope it's not impolite that I "intercepted" your questions twice already :) )

Iron

wahaufler: I expect it's common. In my machine vision system, assumptions about glare and reflections turned out to be some of the biggest "unspoken assumptions"; we had a lot more of both than we first expected.

Iron

QUESTION: Did you achieve your cost goals?
mharkins

ANSWER:  Our customers have or are putting the camera modules into production so I believe so.

QUESTION:  What voltage is that Smart Rear Camera running at?
CurtD

ANSWER:  The module is actually powered from the cars 12 VDC, but the internal voltages varied - 1V, 2.5V, 3.3V were generated internally.

Blogger

@noor811: At the moment HDR-cameras are still a niche product. You have to take what you get. Although digital interfaces will succeed the analog ones in near near future, they are mostly still available for cheap products.

Iron

18 months was my guess.  Thanks

Iron

How is your hardware and software reliability in the field?

Iron

@lvruffino, I agree, with today's technology...

Iron

While working for Boeing, I helped develop the User Interface (UI) for the Space Vision System (SVS), used for robotic assembly of the space station. However, as I suspected, after some research I discovered some hidden design constraints.  We had assumed it would be a typical VGA computer display, but it was only apparent after digging deep in the schematics and documents of the system that the GUI would have to be displayed on an NTSC video signal, severely limiting resolution and requiring use of large plain fonts.  The hardware used also constrained us to an 8 color palette. This was further exacerbated by the severe and harsh lighting environment in the space shuttle.

Lesson learned: be careful of unspoken wrong assumptions. These are hard to ferret out.  Is this a fairly common lesson in the industry, as I suspect?

Iron

QUESTION: What was the development time, overall.
gbilliard

ANS:  From early algorithm deveolopment to field ready prototype about 18 months.

QUESTION:  How to make so small  Smart Rear Camera? what devices they used?
sherlock

ANS:    In the first versions the major components were usually a VGA sensor from Omnivision and CogniVue's CV2201 image cognition processor. 

Blogger

Simon, would an FPGA have been a better alternative, maybe not cheaper?

Iron

Would be nice if Digikey or Design News would broadcast "Virtual Embeded Vision Summit" so we could sit in on sessions without traveling and losing time

 

Iron

Thank you for the answer, Simon! ARM's vector unit is indeed very useful.

Also: Great presentation.

Iron

@Atlant: Thanks, I thought  so, It seems that in library there are minimum intelligence, just collection of patterns.

QUESTION How do you recognize object from background?
mahmood

ANSWER:   The application did not require what is called classification, but determining the distance of the cloest feature.  In particular, it required what is called feature detection and tracking and one of the primary sub-algorithms used to track the features is an algorithm called Good-Features-to-Track.

Blogger

Question:Were you able to use C+ on the embedded platform with simplified data structures etc. or did you have to port it to C, like many embedded platform do not support OOP? Thanks!
ADiewi

ANS:  We used C on the PC, but on the platform it was ported to C.  The master processor was an ARM9 RISC processor and the primary image cognition processor engine was a block called APEX which contains a massive SIMD processor called the Array Processor Unit.  The APU was programmed in SIMD Engine Language.

 

Blogger

@cghaba, there will not be a live stream of the event, but as with last September's Summit, we will be posting the videos of the various presentations in the weeks following the event. See: www.embedded-vision.com/industry-analysis/Video%2BInterviews%2BDemos

Hi Simon, could you please tell me about feature detection, are you able to diferentiate humans, animals, objects, etc?

 

Iron

Simon, why was an analog camera used rather than digital/CMOS?

Iron

Thank you Chuck & Simon

Iron

Thanks Simon, Chuckdigikey, et. al.

Great presentation today and all week.

 

Iron

Thanks Simon and Chuck

Iron

thank you Design news and Digikey

Iron

Thanks Simon, interesting stuff

Iron

Thank you, Chuck and Simon

Gold

thanks, Simon

Good job

 

Iron

Excellent Presentation, Thanks Simon

Iron

Very informative presentation

Iron

Excellent week, thank you everyone!!

 

Thank you.

Tom Scheffelin

Iron

Will there be also live transmission from the Embedded Vision Summit?

Iron

Thanks Si,om.  

Iron

Thank you for the presention Simon

Iron

Thank you very much Simon.

Iron

***** Slide 15 *****

Iron

Anatoliy1086: Is it specific to the field?

It would have to be; imagine, if you will, the difficulty of backing-up towards a "Handicapped Parking" symbol painted onto the pavement. Is that a real obstacle or not???

Iron

***** Slide 14 *****

Iron

@Atlant: Yes, there is definitely a need need for lenses with a "lotus effect". But this will likely harm the optical properties...

Iron

***** Slide 13 *****

Iron

What criteria is for golden image test database? Is it specific to the field?

"Plan many more" -- Boy, ain't *THAT* the truth!

Iron

mrybczn: I have to tell you that in this particular application (car backup camera), road dirt will mean that you can't even get two-nines reliability. The lens of the backup camera on my Volt gets dirty-to-obscuration after *ANY* precipitation, rain or snow.

Iron

Is there a Pedestrian detection library? Does it meet safety guidelines?

***** Slide 12 *****

Iron

Is he allowed to mention some of the customers and products already on (vehicles) already on the market?

 

Can you / How do you get to five-nines reliability?

Iron

and at what volume where the cost goals for?

Iron

Most trouble are the assumptions made which later turned out to be incorrect.

Iron

what was the development effort (staff hours)?

Iron

Did you achieve your cost goals?

Iron

What voltage is that Smart Rear Camera running at?

Iron

What was the development time, overall.

Iron

How to make so small  Smart Rear Camera? what devices they used?

 

Iron

@Atlant: Yes, that's true. But what if a C-compiler is not available for your target architecture ;)

Iron

***** Slide 11 *****

Iron

***** Slide 10 *****

Iron

How do you recognize object from background?

Iron

(The char tool is mutilating comments. I meant to say "What's Cplusplus?")

Iron

ADiewi: That's a tricksy sort of a question; remember that you can write valid C and pass it through a C+ compiler. If the compiler's good, the produced code should be as efficient as code that was processed by a true C compiler.

So, what's C ? ;-)

Iron

A bit late for the live lecture today...

Iron

***** Slide 9 *****

Iron

@SCMorris:

Were you able to use C+ on the embedded platform with simplified data structures etc. or did you have to port it to C, like many embedded platform do not support OOP? Thanks!

Iron

***** Slide 8 *****

Iron

***** Slide 7 *****

Iron

@MazianLab, thanks very much for the slide prompts. I've just emailed Simon and reminded him to announce transitions (and new slide numbers) so I think we'll be set for the rest of the session

I only use chrome for the audio (0 dropouts) and keep my IE for the slides and other lookups during the  lesson.

Iron

High average bandwidth is a much different problem than consistently-low-latency data transmission.

 

Iron

***** Slide 6  *****

Iron

Exactly how difficult could it be ? Gigahertz processors, gigabyte memory, 100's kbps connectrion speeds.

We're much better off with 3kHz POTS telephone audio.

Can't the computer industry get anything right ?

Iron

*****Slide 5   *****

Iron

JoeFromOzarks: Never discount the importance of the "placebo effect"!

Iron

(I don't know if repeatedly pressing play/pause helps, but it makes me feel better.)   :)

@JSP: Try using Chrome as the browser. It seems to work the best.

I can't believe that in 2013, the thin web audio feed keeps dropping out !

Iron

audio working good today, yes!

Iron

Hello from Ottawa ON

Iron

TGIF.  Hello All.

Iron

Hi all - Audio is live!

Hello again from Germany.

Iron

I had about 10 to 20 audio drops at each session. I am prepared to push play button repeateadly. Sometimes it restarts quite quickly, some times not, which is annoying. I use IE9.

Iron

@Atlant I want a tricorder

Gold

Oh

IRON

I am told means Information Online Radio Network

Iron

@fdarocha: The audio is not on yet.  It will be on in a few minutes.  Watch the area above "Our Lecturer" -- the audio player will appear there.

Blogger

78RPM: Someday, I'm hoping for "Dilithium". ;-)

Iron

with chrome audio is steady with ie it keeps braking for me

Iron

Hello,

Is the audio supposed to be on already?

Iron

@rruther2

I guess Silver means that I've participated in a lot of blogs/classes, etc. I did write one Gadget Freak article.

Gold

Hello from Troy, OH

Iron

For me, most of the time, audio gets on & off.

Iron

Good afternoon everyone, Simon is on line.

Blogger

JimmyL: I've gotten the impression (across quite a few of these sessions) that it is a load-related problem, either on the server or on our local systems (and I lean towards the latter because it was awful when I was running a software build in parallel with the session).

StephMcGUBM: I'm actually using Internet Exploder on Win7, although that's not my usual browser. Tuesday, I had no drop-outs running Safari on a Mac.

Iron

Hello from sunny Newport Beach, CA

Iron

@Atlant: Are you using either Chrome or Firefox? They work the best with this system. IE tends to drop audio for many people. I use Chrome and have no problems.

Atlant: Are you using either Chrome or Firefox? The player works the best on them. The audio breaks on IE for a lot of people. I use Chrome and have no problems. However, rarely it does take 2 browser refreshes for the play button to be active at the very beginning of the session.

I had the audio cut-out problem with IE on Monday. I then tried Firefox, same problem. Last I siwtched to Chrome, and so far it dropped only once up to now.

Iron

@Atlant   At the start of the show, I click the play/pause button about 20 times, rapidly turning the player on and off.   I don't have any problems with it dropping out.    {For the pool, put me down for 20 clicks.}  :)

So: Enter the pool: How many times will the Audio Player "cut-out" today?

Iron

Good evening from Iasi, Romania. Here winter has comming back.

Iron

Hello from snowy - SLC, UT

Iron

(He's better than us poor ferrous folks? ;-) )

Iron

Hello from cloudy-rainy Atlanta

Iron

Hello from Edmonton, AB

Sunny after an incredible heavy snow fall.

Iron

@78RPM, What does Silver mean?

 

Iron

Aloha from Montana

Gold

It's a visionary Friday...

@mahmood:  Yes, the slides are available in the usual place:  Look for the link labeled "Today's Slide Deck" just above the chat window.

Blogger

Hello from fffridged Michigan

Iron

slides ready for download yet?

Iron

Hello from Binghamton, NY

Iron

Hello from snowy SE Lake Simcoe.

Iron

Hello, morning everyone.

Iron

Hello from Panama City, FL

Iron

Good afternoon from NJ.  More or less clear skies and a toasty 48 degrees F today.

Iron

Greetings from Chicago!

Iron

Good morning, everyone

Iron

Good morning, everyone!  It's another beautiful day in Walnut Creek, California: www.mtdiablocam.com.  

I am looking forward to Simon's presentation this morning.  It's an updated version of the highly rated talk he presented at the Embedded Vision Summit conference in September in Boston.  And Simon will also be giving a related talk next month at the Embedded Vision Summit in San Jose, California on April 25th.  See www.embedded-vision.com/embedded-vision-summit for info on this free educational event.

Blogger

Good morning from Scottsdale, AZ

Iron

Howdy folks!!  :)  {Five inches of snow last night, it's beautiful!!}

Good Morning from Sunny Boston; TGIF...15 slides. Danke Schon!

Iron

Good Morning from Milwaukee!

Iron

Good moring from portland oregon.

Iron

Good morning from GA

Iron

Good morning from Mobile, AL

You're a day early though!  LOL.

Iron


Partner Zone
Latest Analysis
Factory floor engineers may soon be able to operate machinery and monitor equipment status simply by tapping their eyeglasses.
GE Aviation not only plans to use 3D printing to mass-produce metal parts for its LEAP jet engine, but it's also developing a separate technology for 3D-printing metal parts used in its other engines.
In this TED presentation, Wayne Cotter, a computer engineer turned standup comic, explains why engineers are natural comedians.
IBM's new SyNAPSE chip makes it possible for computers to both memorize and compute simultaneously.
The “Space Kid,” 11, will be one of the first civilians to have his design manufactured in space by NASA, thanks to the City X Project, which inspires kids to think about new 3D-printed inventions that could be useful for humans living in space.
More:Blogs|News
Design News Webinar Series
9/10/2014 11:00 a.m. California / 2:00 p.m. New York
7/23/2014 11:00 a.m. California / 2:00 p.m. New York
7/17/2014 11:00 a.m. California / 2:00 p.m. New York
6/25/2014 11:00 a.m. California / 2:00 p.m. New York
Quick Poll
The Continuing Education Center offers engineers an entirely new way to get the education they need to formulate next-generation solutions.
Sep 22 - 26, MCU Software Development – A Step-by-Step Guide (Using a Real Eval Board)
SEMESTERS: 1  |  2  |  3  |  4  |  5  |  6


Focus on Fundamentals consists of 45-minute on-line classes that cover a host of technologies. You learn without leaving the comfort of your desk. All classes are taught by subject-matter experts and all are archived. So if you can't attend live, attend at your convenience.
Next Class: September 30 - October 2
Sponsored by Altera
Learn More   |   Login   |   Archived Classes
Twitter Feed
Design News Twitter Feed
Like Us on Facebook

Sponsored Content

Technology Marketplace

Copyright © 2014 UBM Canon, A UBM company, All rights reserved. Privacy Policy | Terms of Service