Control Engineering Asia

Sponsored Links

Ads by Google

Add a Comment

» Post A Comment Now!

There are no comments for the article yet.
Rate this Article

Current Rating:
No rating yet

Excellent
Very Good
Good
Quite Good
Poor

Rate this Article Now!

Related Stories

Enhanced Parameter Optimization for ELAU PacDrive - 18 February 2008


ELAU has claimed a significant feature enhancement to its PacDrive automation systems, without compromising stable servo feedback loops for fast commission

Lenze AC Inverter Drive Adds Profibus to Plug-In Communications Options - 18 February 2008


Lenze-AC Tech announces the release of the PROFIBUS-DP communication module, the fourth communication option for the SMVector drive, together with D

Innovation First Announces VEX Robotics Products for 2008 - 15 February 2008


Innovation First, Inc. has announced the introduction of two new products for 2008 - the new VEX RCR Mini robotics platform and VEX 802.11 Wi-Fi


How keen are you to install wireless instruments in your plant?
Very, I see many possible applications
Would prefer to wait for technology and standards to mature
Not at all, I have serious reservations about plant wireless
View results
Ask a Question

Free Magazine Subscription    Printer-friendly version    Email to a Friend

           

The Strobe Solution

-- 1 June 2007

Ads by Google

Dave Appleton reveals the secret to capturing crisp images of fast-moving items with machine vision cameras.

If you are building systems to capture images of objects on a production line, the chance is that you will have to snap images on the fly. If you are not using strobe lighting you are either paying too much for your cameras or you are getting blurred images. Maybe even both. But as this article explains, you can get crisper images without spending a fortune on super fast cameras.

“Blurring” happens when you try to capture a picture of something that is going too fast. The word “fast” is a relative one. Fast in this context is about how quickly something moves across the picture and relates to pixels per second.

This means that if you are zooming in at high magnification (i.e. you have a small field of view) even a small movement can have serious consequences on image quality.


Why blurring happens
Machine vision cameras usually use charge coupled devices (CCD) sensors arranged in a 2D matrix of cells, each representing one pixel of the image. Each cell of the CCD accumulates charge while light is falling on it. If the light is very bright the charge accumulates quickly. If the light is dim it accumulates slowly. If there is no light then no charge accumulates.

The total charge during the exposure is a result of the charge accumulation over the entire exposure time. If the light level changes during this time then the rate of charge accumulation will vary. If there is too much light or the exposure time is too long, the cell will reach the maximum charge that it can hold. Then the image becomes saturated and you lose image details.

An object moving through the field of view will, as it moves, change the light falling on certain cells. The front of the object will cover new pixels as it advances. The back end will expose new pixels as it moves along. These pixels will not charge at a constant rate. Because they register part dark, part light, they will end up indicating gray – an average of the light and dark values.

The seriousness of the problem
As a rule of thumb, the more the object moves during your exposure time, the more likely it is that you will get a blurred image. If the thing you are capturing moves by even half a pixel during the exposure, it can have a serious effect on the measurements.

Many vision systems measure to sub-pixel accuracy using interpolation so the movement can really mess things up. Of course it depends what you are doing. If you are looking to see if a bottle cap has been placed correctly it may not be important. If you are checking dimensions, blurring could render all your measurements meaningless.

As an example, if your camera’s exposure time is 5 ms and you are looking at a field of view of 64 mm at a resolution of 640 x 320 pixels, you will get blurring if the object moves more 0.05 mm in 5 ms, which translates to 1 cm per second. And that’s not particularly fast, is it?

Eliminating blurring
There are a number of traditional solutions to eliminating blurring, each with associated costs:

• Slowing or stopping the object
Sometimes this is possible but usually you become unpopular with the process engineer because you are reducing his throughput.

• Reducing the exposure time
This reduces the window, and hence the opportunity, for a movement to have an effect on the image. But it also reduces the total light level so you need more sensitive cameras with higher gain, or brighter lighting. And cameras with faster exposure times are usually far more expensive.

• Strobe lighting
A far more cost-effective solution, because it creates the effect of a reduced exposure time, even if you are using a low cost camera. The ambient light should be reduced for this method to be really effective, and if this is done, when the strobe light is off, very little light falls on the sensor. With strobe pulses far brighter than the ambient light, the CCD sensor records what happened during the strobed pulse and ignores everything else.

Using strobe lighting – the basics
A frame grabber normally has an output to supply strobe pulses when the camera is acquiring the image. Cognex frame grabbers, for example, have the option to add an optically-isolated interface box which can be used to connect triggers and strobed lights. And USB and Firewire cameras often have an output from a connector on the back of the camera to supply trigger and strobe pulses. Connecting these outputs to a good strobe controller, you can produce extremely bright light pulses with durations as short as a few microseconds.

The better strobe controllers also “over-drive” the LED. In other words, they make use of the fact that short pulses are being used to exceed the ratings of the LED under precisely controlled conditions and so to get brighter light – without damaging the LED. Good strobe controllers can also handle the lighting needs of several cameras firing independently of each other.

Examining some real-world scenarios can help to explain the benefits of strobe lighting.


Case #1 – Fast moving objects
A production line produces 30 parts a second and you need to inspect each part at some point on the line. You’ve done the calculations – 30 parts a second – that’s 33 ms per part. An eternity.

Using a camera with a capability of 70 frames per second equates to a capture time of 14.3 ms per frame, leaving plenty of time for processing even without a buffer. According to the camera specifications, minimum exposure time is 5 ms. You set everything up. You get blurred images.
What can you do?

Well, the first thing that you need to do is to have some kind of sensor to detect when the part arrives. You can use proximity sensors, light switches, encoder outputs among others depending on the situation.

You will need to configure your hardware and vision software to use triggered input for the camera – then use the output to fire the camera to capture the image.

You will also need to configure the software to produce a strobe pulse when the image is captured. Your frame grabber or smart camera will generate the trigger pulse just after the camera starts capturing the image. You will then need to adjust the brightness and pulse duration to get a good image. Reduce the ambient lighting on the object or decrease the camera’s light sensitivity. This will have the effect of freezing the object.


Case #2 – High resolution images of large objects
When inspecting relatively long parts – like electronic connectors or circuit boards – resolution will fall away if the camera is placed far away in order to see the whole part.

Bringing the camera nearer will produce a good close-up shot of a small part of the object – but now you can no longer see the whole thing. The solution: move the object (or the camera), snapping pictures as you go.

Using a servo-driven system allows for trigger pulses to be derived from the encoder output of the controller. Other alternatives include decoding the output of a shaft encoder, or using a trigger comb to interrupt a light beam. With moving items, you need to use strobed lighting to prevent having to pause to capture the images.

The downside is that the vision software needs to become more complex – requiring techniques to stitch the images back together or, at a bare minimum, relating the results measured from each image. This is not a trivial task but it is often the only way to go to get the required resolution.

At Calistra Research Labs we have used these techniques for taking up to thirty snaps of electronic connectors at high resolution. We have a field of view of about 1 cm while moving the object at about 200 mm/S.


Selecting a controller
To realize the benefits of strobe lighting it is important to use a good strobe controller, and these come in several flavors. Most controllers use pulse width modulation to change the intensity of the illumination because LED life tends to be shortened if you apply DC voltage for extended period of time.

Current Control devices work by limiting the maximum current to the LED array. Too much current and the LED will blow up. But this does require tuning the controller to the LED being used. And changing the LED configuration (e.g. from a 60 mm LED array to 30 mm array) will necessitate a change in the current limit – otherwise, the new array will blow up or produce dim lighting. As implied by their name Voltage Control devices work by supplying the specified voltage to the LED array.

Over Driving controllers, meanwhile, deliver too much voltage – but for very short pulses. The short duration of the pulse means that the LED does not over heat or get damaged. Overdriving controllers have the great advantage that they deliver the brightest for any given LED. This is important when you require high contrast between ambient light and the light pulses.

As for intensity, most controllers simply have a knob to set the brightness. So if you just want to set the brightness and leave it, or if you enjoy fiddling about with knobs, a hardware controlled strobe unit could be the best choice.

But for a regular mix of products being inspected, it helps to have a means for the vision computer to control the intensity to preset levels for each part that passes through. In this case, a software interface (RS-232, USB, Ethernet) enables embedding the lighting control into the vision system. It also allows accurate reproduction of previous settings by reusing the values.

Since the trend these days is to build flexible machines that can inspect a number of products, embedding the lighting control means that you can stop “itchy fingers” from messing with lighting levels.

Both hardware and software systems allow you to set intensity and to switch between strobed and continuous lighting.

Finally, for the strobe controller to do its job properly, the pulses must be bright enough to illuminate the object, yet of short enough duration to freeze it. The pulse width must be in the order of tens of microseconds.

If you do manage to select a really good strobe controller, then even with a low cost camera you too may get people thinking that you have faked the images until you wave your hand in front of the lens to prove that everything is live!


Strobe Controller

Lumicon Strobe

           

Free Magazine Subscription    Printer-friendly version    Email to a Friend