ASK THE EXPERTS
More Answers From David Dechow
Principal Vision Systems Architect at Integro Technologies
- Email: ddechow [at] integro-tech [dot] com
- Tel: (704) 636-9666
Who developed the first "standalone" Smart Camera that contained a DSP or Microprocessor for decision making?
By "smart" let's say that we're describing a machine vision device that has the acquisition, possibly optics and illumination, and processing components and software along with suitable industrial I/O to communicate results, all packaged in a single integrated component. One also might have to limit it to devices that were commercially "viable": a few components and custom devices like this were around in the late 1980's or before but were either not marketed or did not achieve broad success. To my knowledge, no manufacturer or person claims to have had the world's first "smart camera". However, Dickerson Vision Technologies introduced a component that I worked with which indeed met all of the definitions above as early as 1990/91 called the DVT smart camera. Likely there's something earlier. (Allen-Bradley had the VIM much earlier but that was a tethered "smart" camera system.) It would be interesting to hear other comments on this.
I am brand new to this site. I have a need for some type of vision that can identify the edges of textiles (mostly all white) such as sheets, towels, Wash Clothes, etc. I would envision two types of robotic arms that would be able to fine (grab) the two ends on a side, not opposite ends, and spread them out. I understand that the vison may now be able to identify white textile. Any help or a company that may have the vison to do this would be greatly appreciated. Thanks Jeff Nichols
Hi Jeff; Machine vision technology is regularly applied to many inspection, guidance, and measurement applications. The thing to keep in mind is that the components that are used are just the tools with which one builds and programs a successful application like yours. That said, the design and implementation of a solution that involves many components is usually handled by an integration partner skilled in that task. While your specific project sounds like one that hasn't been widely installed already, it certainly appears on the surface to be feasible. I suggest that you engage a good integration partner to help you in the specification and if appropriate integration of your system. Since my company does perform those services you can feel free to contact me directly for more information. Also, you will find other similar and excellent integration companies on this AIA website. Many are "AIA Certified", a designation indicating a high level of integration expertise.
I am working on a report on applications of machine vision. I came across terms like "Voxels over Raster Grid" and "Point Clouds", when I was reading about improvements in Machine Vision functions. I am a mechanical engineer and I want to have some basic idea what these terms mean in respect to machine vision. Also how are Embedded Vision and Computer Vision different from Machine Vision.
Hi. I'm curious about the sources you've found relative to your question about Voxels and pixel Point Clouds related to machine vision function improvement. I have not yet seen voxels specifically used in machine vision, but perhaps someone somewhere is moving in that direction. Overall both are just related to the "representations" of 3D spacial scenes or objects and not necessarily related to the functionality of machine vision tools that would process, segment, and locate or measure objects in 3D space. Further, Voxels or "Volume Elements" as opposed to Pixels or "Picture Elements" are a representation more commonly used in things like gaming, movie making, and medical imaging. I think you will find several good webcasts on the topic of 3D machine vision in general here at visiononline.org that might help you in your research. Due to space, I'll leave the other question for another time.
I am looking for a camera that will detect air bubbles leaking from a fuel line submerged in water.
Hello Michael; A common misconception in approaching machine vision technology implementation is that an individual component - like a camera - is the solution for a specific task. In reality, there are dozens of cameras and other components that likely could be used in your application. The key to any machine vision application is in the proper design of the system and experienced implementation of the targeted hardware. Bottom line: The success of your application starts with a competent analysis of what your project needs to achieve and the constraints of the environment in which the project will be implemented, followed by the configuration and/or programming of a reliable application to perform the image acquisition and analysis. If this is not something your company already has experience with in using machine vision, you can find many suitable integration partners on this web site. If you want, feel free to contact me directly for more information. Best regards, David
I was testing a telecentric lens that allowed a fiber optic light source to be inserted in its side. When testing the lens with the fiber optic light source my image got washed out with a grey color, adjusting gain and offset did not help. When I used an external light source with the same lens and camera I got a normal image. The supplier was unable to tell me why my image turned grey with this light. They were able to tell me the lens, camera, fiber optic cable and light source are compatible. Has anyone else experienced this issue and have information on what caused it/how to correct it?
The description of the image as turning "grey" is hard to visualize. The components you describe are standard imaging and lighting products and this illumination technique is common. Can you provide more information? Knowing the camera model, light source color and the object(s) being illuminated would help our understanding. Feel free to respond here or to my email address.
We are a manufacturer that prints on continuous web whether it be opaque but mostly clear. The prints can consist of hundreds of designs and unlimited ink colors since we are a custom made to order company. These production runs could last for short period of time ~10min or anywhere to multiple hours. From my experience with vision and the challenges we produce we are trying to find a print inspection solution where the software could easily learn the print with minimal operator involvement and inspect the material for print defects such as print voids. Some challenges to know is that we do not having any registration marks on our product so it is a continuous web print could be within 54-62" wide. If anyone has any solutions or ideas please let me know, thanks in advance!
Hello Hark; The print inspection application you outline has potential for success, but it would be premature to speculate on the exact details of the required technologies until more information is gathered.. In pursuing this type of solution with machine vision technology, it is critical to fully analyze the requirements, scope, and metrics of the application in preparing a project specification. You will find many useful articles and educational materials here on the Vision Online website, and if you need assistance in the design, specification and integration of your project, you will also find AIA Certified System Integration partners on this website who serve your geographic area.
Hi, I am looking for inspection system which can works for my 2pc alluminium can line to detect the inside defects on the can. Please suggest any ideas
Hello Shintoy; Your first consideration for this inspection should be to seek out manufacturers of Application Specific Machine Vision (ASMV) systems who already have developed turnkey solutions specifically designed to do internal and other can/container inspections and who serve customers globally. Two that come to mind are Applied Vision Technology (www.appliedvision.com) and Pressco Technology (www.pressco.com). Both are AIA member companies and you can find more information about them on this web site. (Note that I am not affiliated with either company and am not commercially promoting any specific product.) You might find other companies internationally who specialize in this type of inspection as well. Good Luck, David
aside from Golden unit, what is the best way to correlate vision systems?
Hello Jay; I believe by correlation you mean the actual validation of your machine vision inspection system's performance. This is a topic that I won't be able to answer in this limited Q&A format, but the short response is that in some applications a set of "golden" parts is the correct way to validate system performance. This works well particularly in cases where the vision task is a "go/no-go" inspection or an inspection for well-defined defects or features. For more subjective inspections, a regular statistical analysis of accepted and failed part samples can provide some validation of performance. For measurements, the vision system reliability should be validated using standard MSA (measurement systems analysis) techniques and statistics like a Gauge R&R, P/T (precision/tolerance) analysis or others. If you'd like to discuss further, feel free to contact me directly. Best Regards, David