Ask the Experts

Objects immersed in water have been imaged clearly. As a scuba diver I see many extraordinary pictures taken in the water.

Certainly the clairty of the water is critical, but I'll assume that problem, if it existed, would be obvious to you. It's also possible for water turbulance to be an issue, but, again, I believe you would see this yourself. Is it possible that you have light reflecting off the surface of the enclosure and reaching the camera's lens, and that would degrade the image.

Motoman sells a vision system for their robot. I recommend you check with them to see if that vision system will work for what you need. Any other solution will cost far more for the engineering. However, be assured that your requirements can be met with a vision system.

More information is needed such as how the inside is currently inspected without machine vision, what is the size of the article, what the inspection needs to check, and so forth. Without that information, I might suggest you consider x-ray imaging.

The question about how you check the dimensions of any object is far too broad. There are many imaging and image anaylysis techniques that can be applied to measurements. Object size, shape, and material make a difference as well as what characteristics are to be measured. The measurement precision required also affects the techniques. If you have a more specific example, then it might be possible to provide you with information.

Assuming you want nominally square pixels (ignoring the dynamic component caused by the part moving during exposure), your scan period is determined by the time it takes the material to move a distance equal to the horizontal pixel spacing.

For you application, 300 dpi is a horizontal pixel spacing of .084 mm/pixel. The speed of the line, 150 M/min is 2500mm/sec. The scan frequency needs to be 29,762 scans/sec (a period of 33.6 usec/scan).

Hope this helps.

There are several good line-scan camera companies. On several projects, I've used the Dalsa line-scan cameras (I have no financial connection to them), but I would consider others, too, like E2v.

The algorithm will most likely need to be custom developed. There is no "true" line-scan software package available; the ones on the market are modifications to area camera programs. While I could ramble on about line-scan algorithms, you gave far too little information to offer suggestions. There are many details needed like what are the sort criteria, how are the grains being presented, what is the production rate, and other system architecture details. This discussion would be too much to handle on this forum. You can contact me directly if you wish to get into greater details.

Steve --

From your question, you obviously know the following that I'll restate as an introduction to my remarks. Considering light sensed by the machine vision system, you have the engineered illumination and the ambient light. Engineered illumination creates the signal; ambient light creates the noise. The problem, then is signal-to-noise ratio. To improve the ratio, you need to either increase the signal or decrease the noise or some of both.

I don't have template language for you as each application is very unique. It's difficult to specify ambient light (e.g., intensity, direction, spectrum, etc.), and this can change for several reasons. I suggest your buy-off test incorporate a provision to compare the vision system's operation under different lighting conditions (e.g., lights off and lights on) and require that the performance not change by more than some modest delta.

Jessica -- Do you know the filter thread for the lens? It should be on the data sheet or available from tech support at Navitar. I would expect it is a standard size. You can order the neutral density filters in threaded mounts from Midwest Optical Systems. I've used their products on a number of applications and been pleased with the quality and service.

As far as I know, there are none commercially available. However, from the artwork generated by the Cognex software, you can order a grid (black on clear mylar) from The PhotoPlot Store in Colorado. Tell Rich Sayer that I suggested him. I have purchased a number of targets from the from large (e.g., 1 meter square) to extremely small.

Daniel -- I am in the San Jose area and would be willing to spend a bit of time with you to help you work out your requirements and possibly identify the components you will need. Please contact me by return e-mail.

Perhaps more explanation of what you are trying to verify would help. If you want to verify the camera settings, then either you need to go into the camera setup program and check all the settings and registers or you need to have a program written as suggested by a prior respondent.

On the other hand, you may want to verify the quality of the image your camera is creating. If this is what you want, I have a white paper in draft form, "Benchmarking the Imaging". I also have a video, also in draft form, that covers the same material. Let me know if you are interested in either of these.