System Integration Questions & Answers

Click a question to reveal the answer. Don't see the information you're looking for? Take a look at the archived questions or ask a new question by clicking the button to the right.

Hi, I was wondering what a bare-bones vision system would consist of? Ideally it would just be a camera, a lens, a cable, 12VDC power (unregulated ok?), a serial port or USB port, and a laptop. Maybe software to control the camera? I don't know when a frame grabber is required and when it is not. What are the different communication platforms? Do most cameras work with any platform? Is RS-232 a platform or does another platform run over serial? Could you recommend software that works with most camera models? Video stream vs. an image stream of continuous pictures? Thanks in advance for your time. I would be interested in attending/viewing a Webinar if you know of such a one. Dan

Hi Dan, Just getting involved with this industry? You will find that there are many layers to this onion and the more you peel the more you'll find. A rapidly evolving industry with numerous manufacturers developing on an almost continual basis. Some trying to lower the entry costs and some pushing the envelope of what can be done in performance. Lots of cameras that are Firewire, USB, GigE, Cameralink, and now HS link and CoaXpress. Each with "standards" for cabling, communications, etc. So, you've come to the right organization to learn about all of this. But, for "bare bones" I presume you mean lowest cost. All depends on what you are trying to accomplish. If it is to learn about vision systems then one ultra low cost approach is to use OpenCV software (free but could violate patents) along with a web camera. I'd be happy to discuss with you further. Dr.Lee@pyramidimaging.com

Answered by:
Rex Lee
CEO/President
Pyramid Imaging, LLC
dr [dot] lee [at] pyramidimaging [dot] com
(813) 984-0125

Hi Dan. Just expanding upon Rex's response, I would strongly encourage you to consider machine vision using an outcome-based approach. What you want to do with the technology drives your selection of components and software. There is no such thing as "bare bones" machine vision . There are components with widely varying capabilities; what is inexpensive for one application may be overkill for another, but might not even work for something else. Keep in mind too that lighting and imaging are the fundamental issues - get that right first for any project. Bottom line: if you just want to learn machine vision, forget about the hardware, and experiment using stored images with free PC-based or even better, "demo" software and components that many product manufacturers would be happy to provide (check out companies who provide smart cameras, systems, and software in the buyer's guide section of this web site). If you have a project, seek out some expert input as you learn further.

Answered by:
David Dechow
President
Aptúra Machine Vision Solutions
ddechow [at] aptura [dot] com
(517) 655-7860, x11

X CLOSE ANSWER

We are going to built a hi res (cca 16MP) full body color scanner with synchronized multiple camera setup. We'll be using a photogrammetric SW (eg Agisoft ) for stitching photographs into a 3D model. Will be used for a biometric and 3D printing purposes. One option is to buy cca 25+ DSLR cameras but I am affraid of "getting lost in cables". Is there any option in MV field, ideally within $100K budget?

Hello Ludvik, We can help you bring GigE Vision over Gigabit Ethernet into your system to ease in cabling complexity. You have the option of purchasing cameras off-the-shelf (perhaps with Camera Link interface - cameras are readily available with this interface in 16MP variants), and use GigE Vision products from Pleora to bring that image data into your PC and imaging software. Alternatively, if you wish to build your own cameras, we can provide products to GigE Vision-enable your cameras - you pick the sensor and build the camera (perhaps with image processing embedded, or not), and we can provide the Gigabit Ethernet interface. Please contact me at john.phillips@pleora.com if you have any questions and/or would like to discuss further!

Answered by:
John Phillips
Sr. Product Manager
Pleora Technologies
john [dot] phillips [at] pleora [dot] com
(613) 270-0625

Dear Ludvik; ISG has 16Mp and 29Mp Cameras with GigE interface that can be used. These cameras have built-in Linux CPU, large FPGA and lots of memory. I am sure you can put together a nice system well under $100K. Drop me an email with more details of your system. Thank you, Kerry

Answered by:
Kerry Van Iseghem
Founder
Imaging Solutions Group
kerry [at] isgchips [dot] com
(585) 388-5220, x101

Ludvik, Please give me a call to discuss, we might be able to give you some pointers to get you going in the right direction. There are a number of options available to you. Andy

Answered by:
Andy Long
CEO
Cyth Systems, Inc.
andy [dot] long [at] cyth [dot] com
888 508 7355

X CLOSE ANSWER

I'm looking for a vision system solution in conjuction with Liquid Penetrant Inspection (LPI) also known as Fluorescent dye Penetrant Inspection. The system concept is to visually inspect parts that have been prepared and subjected to UV lighting. The system is to accept or reject conforming and non-conforming parts by identifying cracks. The penetrant dye remains in the cracks and "glows" under UV light.

Serge - detecting the dye "glowing" under UV light is fairly straight forward. i suspect that the challenge in your application will be efficiently presenting the parts to the camera for inspection. Factors such as the physical size of the parts, part geometry (shape) and batch size, among others, would need to be understood before an inspection solution could be specified. If you would like to give me a call at my office, I would be happy to discuss the details with you.

Answered by:
Scott Gibbs
CEO
Mercator Robotec Inc
scott [at] mercatorrobotec [dot] com
(519) 884-7880

Serge, The only question I have is do you have an old failed part, and some dye, and the dye characteristics. We could run a sample at our machine vision lab, and show you the kinds of results you are going to see. Please feel free to call or email, and we'll see if we can help you.

Answered by:
Andy Long
CEO
Cyth Systems, Inc.
andy [dot] long [at] cyth [dot] com
+1 (858) 537 6541

Dear Serge; Our firm has done a similar project for another customer. This was a custom camera. How many systems per year do you anticipate needed or reselling ? We can quote a custom solution to your exact requirements. If you can provide a detailed specification we can make a proposal to you. Thank you, Kerry

Answered by:
Kerry Van Iseghem
Founder
Imaging Solutions Group
kerry [at] isgchips [dot] com
(585) 388-5220, x101

Mr. Boulet, As Scott Gibbs said below, it is a fairly straight forward application and that presenting the parts is going to be your biggest challenge. Some things to consider are: 1) Lighting - Depending on the angle and surface area it might be difficult to get the exact angle to illuminate the subject. This can be done with a combination of filters and light sources to get the right image. 2) Finding a UV camera is difficult these days. Most manufacturers have done away with their UV line of cameras, but they are out there. If you need help selecting a camera or building a complete system please let me know. I will be glad to assist you.

Answered by:
William Corns
CEO
Wilco Imaging, Inc.
wcorns [at] wilcoimaging [dot] com
(858) 689-9050, x3202

X CLOSE ANSWER

Plan on using a GiGE video cameras to record data for post-event analysis. Image capture rate 6fps; considering using Raw 8 or Raw10 format to minimize total data bits per image. I heard that using the Raw (Bayer filter) formats, I could reduce data needed per image and still allow the recovery of color in post-processing. Is that true and where do I get literature about that capability? These cameras will record space-based experiments without the ability to recover the hardware and within a minimum number of orbits; where the data must be transmitted off the spacecraft Trying to minimize file sizes; my transmission bandwidth is limited (max 6Mbps). Where do I get the information?

Hello Steven, I have found two good links to help explain Bayer filters - http://www.siliconimaging.com/RGB%20Bayer.htm and http://en.wikipedia.org/wiki/Bayer_filter. Certainly one of the benefits of Bayer filters is a reduction in data rate. The GigE Vision specification allows for several types and bit depths of Bayer pixels. Pleora's GigE Vision products can accept Bayer-formatted video, and our eBUS SDK (which can be used with any GigE Vision-compliant camera) can convert from Bayer formats into RGB.

Answered by:
John Phillips
Sr. Product Manager
Pleora Technologies
john [dot] phillips [at] pleora [dot] com
(613) 270-0625

Hi Steven, I see that you have been given a couple of links which explain the nature of Bayer filtering. The data reduction comes from not having to pass the color information from the camera to the capture device. Just the brightness information is transmitted (as with a monochrome camera). Knowing which pixel represents which color allows for color reconstruction post capture. No need to transmit Red, Green and Blue planes. If you use a 1.3m camera in 8bit mode you can transfer one image every 1.86 seconds. I invite you to contact me to discuss your application.

Answered by:
Gene Nolte
Custom Applications Manager
Wilco Imaging, Inc.
gnolte [at] wilcoimaging [dot] com
(619) 977-3797

X CLOSE ANSWER

I am considering to build a stereo system for my research project, and would like to know whether a number of GigE cameras can be connected in practice over the same Ethernet interface and send data simultaneously without significant loss of performance. I know that in Firewire cameras isynchronous transfers are supposed to have guaranteed bandwidth and so on, however I have neither direct experience with GigE cameras nor access to the standard.

The short answer to your question is mostly yes, cameras can send data simultaneously over GigE ethernet. As you probably realize, ethernet protocol is asynchronous with handshaking, potential for collissions, and the ability to retransmit packets. The degree to which simultaneous performance will be degraded depends on the number of cameras connected, the image resolution of each camera, and the frame rate you are using -- in other words, the fraction of the total available bandwidth you are using. If you keep your total bandwidth utilization low, say below 30%, you should experience very few problems. If you are pushing bandwidth limits, consider a high-performance interface card with multiple ports -- each dedicated to one camera and serviced with its own on-card IP stack.

Answered by:
Perry West
President
Automated Vision Systems
perry [at] autovis [dot] com
(408) 267-1746

Hello Joaquin, Perry West 's comments below are correct of course.. Another approach is that if you really want an arbitrary number of cameras to capture simultaneously, I would consider attaching them all to a common trigger. Some cameras have hardware triggers and internal storage that can then be read by your processor on your schedule. This eliminates any image capture timing differences. This approach may be more expensive, but it guarantees simultaneous capture. I hope this helps, David.

Answered by:
David Wyatt
CEO
Automation Doctor Incorporated
david [dot] wyatt [at] automation-doctor [dot] com
(574) 344-7936

In addition to the great answers posted before mine, I just wanted to comment on your mention of the standard. The GigE Vision standard does not specify performance targets/minimums for compliant cameras, software, etc. - it's up to the developer/implementer to create the best implementation that they can. That is to say that in your situation, the number of cameras that can be connected to the same Ethernet interface may not simply be limited by the bandwidth available on the link, but by the performance of the GigE Vision SDK, along with any driver employed with that SDK. It's not unheard of for the performance receiver-side (PC) implementations to vary by quite a bit.

Answered by:
John Phillips
Sr. Product Manager
Pleora Technologies
john [dot] phillips [at] pleora [dot] com
(613) 270-0625

Joaquín Ferruz, GigE was built primarily for the reason you described. It allows multiple cameras to be connected to one switch and have a consistent data stream going to the PC. If you would like to discuss further or get a demo of Baumer cameras please let me know.

Answered by:
William Corns
VP of Sales
Wilco Imaging, Inc.
wcorns [at] wilcoimaging [dot] com
(760) 994-1125

X CLOSE ANSWER

I am researching the possibility of using vision sensors (camera/laser) to scan components which are worn out, comparing the acquired image to a CAD image of the part as it should be, and then using robotics to weld/clad the missing areas to original dimension. Are there any existing off-the-shelf products that do this? Any contact with experts in this area would be appreciated. This is obviously going to require integration between image acquisition, CAD software and welding automation and robotics.

Jason - There are off the shelf components that will do the various steps in your process. Faro 3D laser scanning would be able to measure the surface topography of the part and build a CAD model from the data. RobotMaster or Robotworks software can generate robot welding paths from a 3-D CAD model. I am not aware of an off-the-shelf system that would completely automate your process. The components described above (or other similar products) could be integrated together to semi-automate your process. My company has experience building machines for hardsurfacing applications that incorporate machine vision and automated welding.

Answered by:
Scott Gibbs
CEO
Mercator Robotec Inc
scott [at] mercatorrobotec [dot] com
(519) 884-7880

Jason, We specialise in 3D scanning technologies, and Robotic Integration. We are an integrator based out of California, but you're list of wants matches our abilities to deliver. I'm going to suggest the following, we like to get a sample before any engagement occurs, to be sure that we are able to perform. We would then run this through our machine vision lab, and then rather than do a fancy proposal, we'll show you what you get with 3D images of your scanned part, and if you send a CAD drawing too, we'll show you the comparison. Please contact me if you're interested. Andy

Answered by:
Andrew Long
CEO
Cyth Systems, Inc.
andy [dot] long [at] cyth [dot] com
+1 (858) 342 6795

Dear Jason; Go to http://parlec.com/pages/presetting They may have a suitable solution. Regards, Kerry

Answered by:
Kerry Van Iseghem
Founder
Imaging Solutions Group
kerry [at] isgchips [dot] com
(585) 388-5220, x101

Good morning Mr. Bond, our 3D machine vision library SAL3D provides functionalities for acquiring a cloud of points using a camera laser system and then compare/align it to an imported CAD file. If the whole system is correctly calibrated it would return real metric coordinates that could be used by a welding robot. You can find a press release regarding these tools at: http://bit.ly/ulQ2e3 Please contact me at dsabate@aqsense.com and I will forward your project to our technical department so that they can analyze its feasibility. More information, specially on the objects shape, would be very useful.

Answered by:
David Sabaté
Marketing
AQSENSE S.L.
dsabate [at] aqsense [dot] com
34 972 183 215

X CLOSE ANSWER

In the Camera Link interface, is the image data transfered in Bayer format, or RGB?

Either one. To reduce transfer bandwidth, most cameras using a bayer filter sensor send the raw Bayer data, which is transmitted the same as a monochrome image, and any bayer decoding gets done either in board hardware, or via post-processing on the host computer. Either way the decoding step is not part of the Camera Link specification so any method for enabling and performing that will be framegrabber-or application-dependent. 3-CCD cameras and some bayer cameras send as RGB data. Usually it's 24 bits per pixel, 1 or 2 taps, and most FGs have standard ways of handling that. Less commonly you'll get greater pixel depth, e.g. 10 bits per pixel as 32-bit data. In that case the MUX at the FG end may be different depending on the FG -- for example it might be the MSBs of RGB in the first 3 bytes, then the LSbits as 2R 2G 2B 20. Feel free to email me if you have any other questions.

Answered by:
Randall Henderson
Senior Software Engineer
EDT
randall [at] edt [dot] com
(503) 690-1234

Hi Mike, Base camera link uses five low voltage differential signaling (LVDS) pairs to send up to 24 bits of pixel data in a serial fashion. Medium and Full camera link use more LVDS pairs to send 48 and 64 bits of pixel data respectively. So, Camera Link really doesn't care how pixels are formatted on the CCD. I hope that helps.

Answered by:
David Wyatt
CEO
Automation Doctor Incorporated
david [dot] wyatt [at] automation-doctor [dot] com
(574) 344-7936

Mike, It really depends on the camera and the way that it processes the image. Camera Link is just a way for devices to communicate. Hope this helps.

Answered by:
William Corns
President
Wilco Imaging, Inc.
wcorns [at] wilcoimaging [dot] com
(760) 994-1125

X CLOSE ANSWER

I need advice on choosing the best software to interface with my GigE camera. I have purchased a Dalsa Spyder3 GEV (linesan 2k mono pixels) and Dalsa Sapera Essencial SDK. I am unable to run my application at the camera's high speed (36,000 lines/sec) without latency issues. What would you recommend for a software interface to my camera? I would also like to stay as generic as possible to facilitate using other cameras. Recommendations for PC requirements and Operating System would also be appreciated. I am a C++ programmer and simply want an API that receives pixels from the camera and puts them in an array for me to process. Thanks!

Frankly, you are really near the limit of the capacity of the GigE link. The real data capacity for GigE is about 80 megabytes per second and, at full speed, this camera generates 72 megabytes per second even with no gaps in the data. This means that to have any chance at all you must use the GigE link as point-to-point (nothing else in the path, probably not even a router or switch) and you should probably use a GigE card that includes full service of the Ethernet protocol so that it is never necessary to wait for the computer CPU. You really need to switch to Camera Link to avoid these problems.

Answered by:
David Gilblom
President
Alternative Vision Corporation
dave [at] alt-vision [dot] com
(520) 615-4073

There are a few vendors that sell drivers for GigE cameras that are quite efficient. I'm most familiar with National Instruments' Vision Acquisition Software, which includes IMAQdx. IMAQdx includes drivers for both IEEE 1394 (FireWire) and GigE Vision cameras. The support for the GigE Vision cameras actually is a two piece driver. One will work with any GigE NIC (the filter driver), and it sits on top of the driver for that NIC and filters packets out that contain camera data. The other driver is for any Intel Pro 1000 NIC, and gives you much better performance on your acquisition. I'd recommend going with this software (available on NI's website) as it will give great performance and works with standard off-the-shelf NICs. On the hardware side, make sure the NIC is an Intel pro-1000, and ensure that you have a few gigs of ram and a decent speed on your FSB. If you're trying to stream this data to a hard-drive at this rate, you might need a raid array.

Answered by:
Matthew Slaughter
Vision Specialist
Cyth Systems, Inc.
matt [dot] slaughter [at] cyth [dot] com
(888) 508-7355

X CLOSE ANSWER

Hello. I am interested in the minimum requirements for tracking humans walking out of a store in a shopping mall. I would like to be able to determine the number of shoppers that walk out of my stores with a shopping bag in hand. This is industry research data I am trying to gather from a retail environment where many people may be walking in view of the camera. Specifically I would like to know the number of people that walk in to my store, out of my store, and the presence of a clothing store shopping bag in hand. Ideally a small standalone camera could either do all processing or provide video data for off-site analysis. Solution should be small and require only a network connection to provide video out or a the various people counts we desire. Solutions are also ideally under $200 for the camera-location installed component. A small systems board nearby or integrated with the camera system would also be suitable. WiFi transmission of data out is also ideal.

Cyrus, The cost constraints seem to be quite tight, but I was wondering what volume of systems you are looking at? This obviously has an effect on these types of systems.

Answered by:
Andrew Long
Senior Director
Cyth Systems, Inc.
andy [dot] long [at] cyth [dot] com
(858) 342-6795

X CLOSE ANSWER

I'm looking to inspect flat sheet metal patterns. Ideally i want to take a picture with the sheet metal pattern on a light table, a software to detect the edges then compare this with a dxf or 2D electronic drawing of the part, complete with GD&T. Are there off the shelf solutions for this, or software which imports pictures and gives the user the ability to generate a report for the part?

MVTec's software package, Halcon, can do this in 2D and 3D. You can get more information by going to our site www.pyramidimaging.com and selecting MVTec under the Manufacturer's area. We can get you started with a 30 day free evaluation copy of the software. It is meant for programmers but the HDevelop tool is available for rapid prototyping and can give non programmers an idea of what operators can be used. We can also provide software programming as a service for you. The task that you are outlining above should not be too complicated. Please contact us at your earliest convenience. Feel free to forward some images if you have them. Cheers!

Answered by:
Rex Lee
CEO/President
Pyramid Imaging, LLC
rlee [at] pyramidimaging [dot] com
(813) 984-0125

Nasir, We have a solution similar to this that could be tuned for your purpose. We have a tool scan inspection system, where tools are scanned on a light table, and then CAD data is produced for cutting the foam that tools are placed in inside the tool boxes. If you would like to discuss further, please give me a call, my info is below.

Answered by:
Andrew Long
CEO
Cyth Systems, Inc.
andy [dot] long [at] cyth [dot] com
(888) 508 7355

X CLOSE ANSWER

I wonder if there is machine vision system integrator that developed an automated inspection system to inspect printing dial for vehicle instrument cluster or if there is anyone who has experience with this type inspection system

This has been done a number of times. Please let me know if you still need assistance and I can put you in touch with someone who could help you. Thanks Kerry

Answered by:
Kerry Van Iseghem
Founder
Imaging Solutions Group
kerry [at] isgchips [dot] com
(585) 388-5220, x101

X CLOSE ANSWER

We have a bad ECP QUAD MULTIPLEXER by ULTRAK--installed (5) years ago. It appears this multiplexer is no longer produced. What is the replacement model number?

A quick web check shows that ULTRAK made general-purpose security camera multiplexers. You can still find many of them either used or refurbished. Probably your application is security not machine vision. If it is machine vision then this is the wrong product anyway and you should seek advice about your application in general. However, if you're looking to replace this for a security camera installation, it is a generally simple RS170 multiplexer and could be competently replaced with a number of current products by any qualified security product distributor. Check for a distributor in your area. David

Answered by:
David Dechow
President
Aptúra Machine Vision Solutions
ddechow [at] aptura [dot] com
(517) 272-7820, x11

X CLOSE ANSWER

View category list

Current Category:
System Integration

Ask Your Question

Close Without Submitting

Fill out the form below. Required fields are marked with (*).

Choose a Category: *

Type Question: *

First Name: *

Last Name: *

Uncheck box if you want your LAST NAME hidden from our Experts, it will not be displayed on the website either way.

Business Title: (will not be revealed)

Company: *

Uncheck box if you want your COMPANY NAME hidden from our Experts, it will not be displayed on the website either way.

Email: (will not be revealed)

To help our Experts provide the best possible answer, please provide the following information (this will only be revealed to our Experts, it will not display on the website).

My Company Is: *(please select only one)

My Company's Primary Product or Service is: *

My Company is Located:

City: *
State/Province: *	(Type NA if not applicable)
Country: *

I agree that by accepting free resources from AIA that I may be contacted by members of the trade association. AIA sends out periodic correspondences about Association updates and news. This is considered a part of the business transaction you have chosen.