OT: line scan cameras

Go To Last Post
19 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quite a bit off topic, but at least Atmel is one of the leading manufacturers of digital line scan cameras, and I suspect that I'am not the only one in this forum that have worked with digital imaging systems.

Anyway, I need a digital image capturing system mounted on a small (100kg or so..) vehicle traveling along a line of crops on a field. The image capturing system have to take a digital image of the entire row of crops.

One way to do it is to use a standard CCD camera and take one image every 20cm or so. This is how it works right now.

Another, much more neat solution, would be to use a digital line scan camera. I could easily get one pulse every 0.1mm or so from an encoder on the wheels of the vehicle. If I make the line scan camera trigger on every pulse from the wheel, I'd get one line of pixels every 0.1mm. This would certainly result in a very loooong ultra-high quality image of the entire row of crops IF the ground the vehicle travels on is perfectly flat. Well, it's a field, it's NOT perfectly flat, my question is, can a system like this still be used without any significant image quality loss due to "imperfections" on the ground?

The image will be used for image processing to measure various aspects of the crops, and the imagequality has to be quite good.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Wow, what a panorama! I can see other applications, like scanning reels of film.

For image quality, you're mounted on a vehicle weighing only 100kg. The wheelbase and wheel diameter will probably be more important than the suspension for long term stability. However, short term (such as the size of an ear of corn), the platform may appear quite stable. Some damping might be necessary if the vehicle's motor vibrates everything. If you're measuring plant height above the terrain, the soil may need to be in view of the camera.

Maybe you can measure the system motion. By considering the exposure time and the smearing you see in the existing images, you may be able to calculate the stability of the platform. But as a practical matter, what software would you use to view these panoramas?

admin's test signature
 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

It's custom software, as well as custom hardware.
The 'panoramas' will be analysed in realtime on the vehicle, as well as stored on a harddrive for further analysis.
The motors running the vehicle are electrical.
I'am worried about the ground not beeing very flat... what happens if one wheel runs over a small stone or pile of dirt? The motion of the vehicle would not be very smooth.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

You are trying to move along 1 axis (z), and you would therefore have to compensate for motion in the other 5 axes (x,y, and the three rotation axes). Compensating for up/down (y) axis motion is possible in software, but at first glance the rotations are a complete pain.

Low tech solution: rails.

High tech solution: mount the image sensor on a damped, actively stabilised platform.

Sean.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I agree that compensation for rotations is the problem. I don't think that movements in x or y directions will be large enough to be much of a problem.
The vehicle will probably be controlled by a PID regulator to drive in a straight line along the row of crops.

Perhaps some kind of gyros can be used to compensate for rotations in software, but that adds alot of complexity to the hardware, and the idea was to make the hardware simpler by moving to linescan instead of CCD.

Both gyros and stabilised platforms are probably more expensive than a CCD camera with framegrabber.

admin's test signature
 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hello,

Compensating for rotation might not be that hard actually, if you take a note from RC helicopters.

They have very small (and somewhat cheap) gyros on them to compensate for the rotation of the main blades. They have an output for a servo that would normally be connected to the tail blade pitch, but instead connect that to a platform with your camera on it.

Here is a nice one for example:
http://www2.towerhobbies.com/cgi...
It is actually dual-axis gyro, you might not need that....

http://www2.towerhobbies.com/cgi...
is a list of all the gyros they have.

Here is a fairly powerful servo, as long as the platform isn't too hard to move: http://www2.towerhobbies.com/cgi...

This may or may not work for you, but though i'd mention it!

-Colin

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

as long as the platform isn't too hard to move

The platform should only need to include the sensor and the imaging optics. Everything else can be on the unstabilised portion of the vehicle.

If the expected disturbances are likely to be small and high-frequency, passive stabilization using a massive table suspended on damped springs may also be sufficient. Making the whole vehicle's suspension as soft as possible might work, but you would get problems with "wallowing" due to low frequency disturbances (such as stopping and starting).

The following link may be of interest, especially the discussions on "vehicle smoothing factor" (VSF): http://www-dev.ri.cmu.edu:8080/p...

I still think that rails are the best solution, if the track is known beforehand. After all, it's what they do in Hollywood for long tracking shots.

Sean.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Well, the track is known beforehand: Along all the rows of crops on the entire field.
So rails are not a realistic solution.

Colin: $280 is not very cheap. And that's just the cost of the gyro, servos are not that cheap either. Especially as I can get a CCD camera for almost the same price.

Only reason I'am considering a line scan camera to replace the current CCD camera based system is to make the hardware simpler and possibly cheaper.
(A CCD camera is quite expensive, and the framegrabber buffer has to be large enough to hold an entire frame = lots of SRAM)

A line scan camera is much cheaper, and I don't need to build much of framegrabber to use it. But if I need costly and complicated suspended and gyro controlled platforms then the line scan camera idea is even more complex and more expensive than the current CCD camera system.

Thanks for your input on the subject, but if rails or gyros are the only ways that might make this work, then the current system is preferable.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I think we got off track here (pun intended).

If you have 300 cm wheels for example, whatever terrain they can manage will result in fairly low frequency disturbances. So, while passing by an ear of corn (15 cm?) you should still get an excellent picture. If you're photographing watermelons or pumpkins, they might have a distorted shape - but the overall size, color, texture, etc should still be perfect. If you're counting beans or strawberries (or just checking color), the terrain probably won't even be detectable.

It really depends on what information you plan to get from your survey.

Note: Your memory requirements really aren't going to be reduced, all the data from a line every 0.1 mm needs to go somewhere. I suppose with a continuous flow of data, it could go directly to disk.

Have you thought about your optics? The CCD camera is usually a very small array, but the line camera might be a long thin affair which would be more complicated to focus, especially if you need to focus near and far objects at the same time.

Finally, if you really really need a straight picture, a straight line (a white string? distant power line?) down the row could provide a reference that the software could use to realign every line image to that reference.

admin's test signature
 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Look at http://www.gis.leica-geosystems....
They use line sensors in aerial photography for mapping and photogrammetry. So it seems to work, but with stabilized platform, GPS, INS and *massive* postprocessing of the acquired data.

admin's test signature
 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

with some cameras you can make a panoramic picture out of a lot of pictures, so a way to go is to tke a picture for each meter (some overlap) with a digital camera in pano. mode a use the pc software make one big picture.

Jens

admin's test signature
 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hello,

Colin: $280 is not very cheap. And that's just the cost of the gyro, servos are not that cheap either. Especially as I can get a CCD camera for almost the same price.

Yes I guess it depends on what this is.. at first I thought it might be one of those company project where it will end up selling for a lot anyway, so $280 isn't that much compared to the total cost...

What sort of price-range are you looking at anyway?

I can't think of anything that would be much cheaper... short of doing some iimage correction by referencing it to some point (such as the horizon, or where the ground meets the crop, or an artificially inserted line) if you do need to neutralize the disturbances.

-Colin

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Well, the track is known beforehand: Along all the rows of crops on the entire field.
So rails are not a realistic solution.

Aha. Yes, rails are out then. The original post said "a row of crops" singular, so rails might have possibly been OK for that.

So you are forced into the difficult art/science of either image processing, platform stabilisation, or ballooning...

Sean.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

mike:
Well, with 300cm (3m) wheels, the vehicle will certainly be very stable ;-)
The current crop of intrest is sugarbeats, but the idea is to use it on other crops as well.
The goal of the research project is to develop image analysis algorithms that can measure many important factors such as weed, insect infestation, crop quality...
The guys who are working on the image analysis want very good quality images and a high performance image processing system, so it's my job to give them good quality images and a high performance DSP hardware platform.

Colin:
At the moment this is a research project, cost is not a big issue.
I only meant that I don't see any benefit in going from a CCD camera based system to a line scan camera system if it requires the addition of mechanically complex gyrocontrolled suspended platforms, that are both more complex and more expensive than the current solution.

Sean:
Sorry I wasn't totally clear in the original post.

Yeah, ballooning, hmm.. a really intresting idea, but I don't think it's a working option, as I said before, the images have to be of very high quality. At the very least 100dpi. Hmmm, I heard US and russian spy satellites are approching that resolution? Now that would be a very neat approach :-)

As it works today, I've built a framegrabber controlled by a DSP that grabs a image every 20cm or so, the DSP then puts all the images together and do some analysis regarding some aspects of the crops. A harddrive will be added soon to be able to store the "panorama images" for later analysis.

I guess I'll stick to this configuration.
This is/was a basic prototype system, I'll soon get started on a revised version with a much faster and more capable framegrabber, as well as a camera with higher resolution and a whole pile of other features. When I'am done they will probably be able to run the vehicle by remote control from their university office :) Before I get started I wanted to make sure that the current configuration is the best way to do it.

I've looked into 166MHz 32bit wide synchronous SRAM's, they exist in large densities (512kB-2MB) at a decent price. They are used as cache memory for Pentium and PowerPC processors. I think they are the perfect candidate for my new high performance doublebuffered framegrabber.
Another alternative would of course be SDRAM, combined with a SDRAM controller and a FIFO, and perhaps even a CPU dedicated to managing imagetransfers.

Does anyone know if 2.5" harddrives use 5V signaling or 5V tolerant 3.3V signaling? I'd like to avoid "infesting" the 3.3V powerplane with small 5V areas.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Does anyone know if 2.5" harddrives use 5V signaling or 5V tolerant 3.3V signaling?

Certainly, the ones I have played with use 5V.

At least they run off a single voltage (the motor drive supply is separate, but still 5V), unlike the big drives which require 12V as well.

Sean.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Yes, I know they run off a single 5V connection, but that doesn't mean the electronics is not 3.3V. They could have an on board LDO regulator.

admin's test signature
 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Forgot to write my nick i the prev. post... :)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Where do you live ?
I know a company in denmark making a system like that.
And sugarbeats are only normal in that area so...

Jens

admin's test signature
 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Sweden.
Yes, I know about that company. They use a combination of laser and imageprocessing to do mechanical weed control.