There is enough hub ram for an entire image... assuming that the average bit depth of a pixel is around 2.5. A raw image coming straight out of Hub Ram with 2.5 bit depth wouldn't be very pretty. the number of available colors is 2^bitdepth... so you are looking at raw images with around 6 available intensity bands, a little better than a binary image. So, algorithms that use binary thresholded image logic should be fine... and there are a lot of applications for those.
I once measured the average bit depth of a rather large series of 8 bit images and found it to be around 6.1.
For example... let's say you want to do position and orientation measurements... and color coding of the part is possible. All you have to do is put a filter either matching or contrasting with your color coding and then object detection and position becomes a thresholded binary image process. LED's can sometimes replace filters. In pathology... a variety of stains can be used to contrast particular parts of a cell... that process almost converts counting processes to a binary thresholded object detection problem.
Shape measurements are required to confirm these count... but sometimes these can be sampled measurements.
You said ..."There is enough ram in the prop for a propcam frame I gather."
That seemed like enough of an invitation to me[noparse]:)[/noparse]
I know that you know this stuff... but I thought I would elaborate for others who might be wondering. For those interested in full color panoramics from their super prop bot... it would be a little on the slow side, but the Propcam will be able to acquire all of the data you could want... buy flipping various filters... but the easiest way to composite the images right now would be to download the raw data to your computer and point ImageJ... from NIH... at your data and then run a plug-in... that I promise to write.
Today I took a rather expensive plunge in the PropCAM project, and now I can explain what's been going on. A few weeks ago DigiKey informed me that the image sensor chip I was planning to use was discontinued by the manufacturer, but that they had 1500 remaining in stock. In checking with the manufacturer, I discovered that they had zero inventory, and that DigiKey's stock was pretty much it, worldwide. What happened is that the fab line that made the chips was shut down and, rather than retool for a different fab line, they just decided to can that portion of their product line.
So I was left with a dilemma: start over completely with a different chip, or do a "limited edition" using the current one, while I searched for its replacement. In the end, I felt that the current chip offered enough unique advantages that it was worth plowing ahead with it and that it would fill a niche, even when its replacement was found. Plus, I'd already put so much work into the development that I couldn't bear just to scrap it.
So today, after much hand-wringing, soul-searching, and attempts to locate additional inventory through brokers, I purchased DigiKey's entire stock of 1500 (and I'm old enough to remember when the amount I invested would've bought a house). So now I'm really committed — or, perhaps, should be committed! In the meantime, my efforts to locate additional inventory will continue; but at least there's an assured supply for 1500 PropCAMs, version 1.
Whoa, Phil, that's scary -- but I'm sure you can count on us to make that the right decision ;-) Thank you.
Does this mean PropCAM's almost ready, or has the stock situation just forced your purchase early?
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Cheers,
Simon
BTW: I type as I'm thinking, so please don't take any offense at my writing style www.norfolkhelicopterclub.co.uk
You'll always have as many take-offs as landings, the trick is to be sure you can take-off again ;-)
I suspect it will be a self-limiting matter, so I see no value in rationing. Those in need of large quantities or a continuous supply will want to evaluate their future demand versus supply before making the plunge. But people have to do that anyway — especially when they're dependent on large manufacturers who can yank a product line with seeming caprice. At least now, with a secure but finite supply, potential users know ahead of time what the risks are. It could've been much worse: I might've made the first batch of 300, then discovered I couldn't get any more.
And it won't end with PropCAM1. As soon as I find a suitable replacement sensor, development will begin on PropCAM2. I read that Micron is reorganizing its imaging division to focus on low-res sensors. This is good news for all of us, and I'm anxious to see what comes of it.
From a technical standpoint, the PropCAM has been ready for quite some time. There's an adapter board that Parallax is having made which breaks out the 2x6 2mm header on a daughterboard (e.g. PropCAM) to a 0.1" SIP. Until the MoBoProp motherboard is ready, most PropCAM users will need this adapter. The adapter board (DB-Expander, Parallax p/n 28303) will be available soon, I'm told. There's also an extension cable, which just came online this week here. This will permit separation between the PropCAM module and the DB-Expander.
Now, with all the necessary accessories becoming available, and with lenses and sensors in hand, there's nothing in the way of getting some boards made!
Well now we know what sensor you are using . Micron may already have a usable alternative in the MT9V022. It is 752x420 but with 4x4 binning you get 188x105. It also has an interesting feature for linking two sensors for stereo vision and they have both monochrome and RGB versions as well. It also seems to support a serial bus using 8/10-bit LVDS, I wonder if we could interface that to the Prop. I will be quite happy with V1 but am already looking forward to V2.
Any availability time frame now that you have your sensors on order?
I've been looking at the new Micron sensors, and they really do have some nice features. But all the ones I've inquired about that have the TrueSNAP global shutter are quite expensive. This is puzzling, since they seem to be aimed at the cost-sensitive automotive industry.
bambino,
The MoBoProp board will not be necessary in order to use the PropCAM. With the DB-Expander and optional extension cable, the PropCAM can be used with virtually any Propeller board. You will want to keep Ports 0-7 available for it, however.
Once the MoBoProp becomes available (no timetable yet), it will likely be the platform of choice for most PropCAM users, due to its small size and the availability of extruded housings for the complete system.
Phil, I have two of the MT9V032L12STM arrays from Micron, they have truesnap and can also do 4X4 pixel binning to reduce resolution.
They cost 11GBP each, that's probably about $20 probably less.
As an aside, Omnivision do some 5megapixel arrays, one site has them for over 100GBP each, but another wants 9.60GBP if you buy 34 of them. They just expect developers and scientists to pay it.
They seem to be a good bit more than that over here: $23.71 each in 96-unit lots. But still, a worthy candidate for the PropCAM2 and less money than I was led to believe earlier.
Well Phil although I want to record lots of frames to RAM my first step is getting the camera to communicate with the propeller at all and then perhaps take a few images for good measure. I'm wanting to get a PropCAM partially because it is a complete working unit and partially because of the software it will come with, I might be able to port some of the functionality over to this array. That might give you a head start if PropCAM2 ends up using this array or similar.
Graham, Are you going to use the serial or parallel interface on the Micron units? You mentioned that you have two of them, are you going to try stereo vision? As you said, I am going to start with the PropCAM because it is a "working" system but I want to play with the Microns when I have more experience.
I'm not doing stereo, I bought two because one will be soldered onto a board on its own to be messed with, the other hopefully will be used to form a proper camera unit.
There's one fly in the ointment, if the Micron sensors are adopted for use with the Propeller. According to the Micron website, datasheets for their newer sensors are available only under NDA. This would imply that any interface details revealed by the datasheets and not by any publicly available information need to be kept confidential. Based on the public info I've seen, this would include the addresses and uses of all the configuration registers. It would be nearly impossible to shield Propeller users from this information, due to the open-source nature of Propeller software. When I broached this issue with a Micron rep, he suggested that the end users of such systems would also have to sign NDAs to have access to information revealed by the datasheets through the software source code. I guess if you're designing rear-view cameras for cars, the NDA thing isn't such a big deal. But if you want to build a camera that someone can actually understand and program themselves, the NDA requirement presents some challenges. And that includes any discussion of programming details here in the forum.
There has, of course, been much discussion here about how to obfuscate Propeller code. Even if I weren't opposed to such measures philosophically — which I am — it would be a risky path to take, given the likelihood that someone would be able to crack the system, exposing the original programmer to liability for damages should the cracker publish his findings. (I haven't actually seen the NDA yet. Perhaps there's a loophole that employs the phrase "all reasonable measures".)
In any event, it will be interesting trying to weave through this legal thicket to come up with a camera based on these chips.
I've been thinking about this Phil and I'm struggling to work out what it is they are trying to protect, is it the interface method or the way the camera works full stop? Answering that question might allow a partial data sheet to be produced containing purely interfacing issues or at the very least allow a code library.
Personally I think the whole thing is ridiculous, if a competitor or a potential cloner wanted to get the data sheet they would just get someone to sign the NDA and then disclose the information to them. When they see the clone device available what are they going to do, call every one who signed the NDA into a room and not let them out until someone owns up?
They need to get a grip, they don't do it for their other products. At least Micron's partial data sheets have enough info to get a decent idea about the chips capabilities and the interfacing issues, Omnivision's just say how good they are and that's about it.
Alternatively you add a CPLD or something to produce an interface interface. So silly though I guess it could be ascii based and quite user friendly.
It's too bad something like a CPLD would be necessary, but it may well be. I guess, as long as there has to be one, it might as well be put to other good uses. For example, while compressing ten bits af A/D output into four, a lot could be done with custom response curves. I had actually contemplated doing something like that with the PropCAM, but with the limited board real estate, it would have entailed putting SMDs on both sides of the board. And that, in turn, would've necessitated a four-layer board just to have an unbroken groundplane behind the lens holder that light can't get through. Such a move would've made the PropCAM much more expensive than I wanted it to be.
Still, it's a shame to offload functions from the Propeller that it could easily handle just to accommodate Micron's paranoia.
BTW: I type as I'm thinking, so please don't take any offense at my writing style www.norfolkhelicopterclub.co.uk
You'll always have as many take-offs as landings, the trick is to be sure you can take-off again ;-)
The major parts are in hand. I made some last-minute mods to the board layout just today to accomodate an LED strobe mezzanine board. I'm trying to make the same strobe compatible with a linescan imaging board for the MoBoStamp-pe that will debut about the same time. I really hate to develop products in isolation from one another, so I try to make everything I design either interoperable or interchangeable with other stuff. This often entails a lot of fussy fiddling. So one more quick proto round -- masked and screened this time -- then off to production.
Also there's a small but necessary adapter board that's been held up in production waiting for a connector. This is the DB-Expander that converts the 2mm header on the PropCAM to a SIP so it can be interfaced with the current crop of Propeller boards. So, meanwhile, I've been busy getting other projects out the door. This includes some new stuff for Parallax as well as fishing lure molds for a tackle company I'm associated with.
I know this all seems interminable. What was it? Ten months ago this thread was started? Cripes! That's beyond embarrassing. But progress is being made; and it will get done. It has to: I've got too much time and money (> $15K) invested in it not to!
As always, I do appreciate everyone's patience and forbearance!
I think we all know you are trying your best and once it is available you can turn the tables, "have you used it yet, have you!!" [noparse]:)[/noparse]
I think Graham's nicely hit the nail on the head there!
No worries Phil; I just know this is gonna be GREAT when I get it
Keep up the good work
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Cheers,
Simon
BTW: I type as I'm thinking, so please don't take any offense at my writing style www.norfolkhelicopterclub.co.uk
You'll always have as many take-offs as landings, the trick is to be sure you can take-off again ;-)
Phil,· Where's my propcam!!·
If we all focus our thoughts on those would thwart your efforts, would it help?
Perhaps, like in the movie Scanners? http://www.imdb.com/title/tt0081455/
fish
Comments
There is enough hub ram for an entire image... assuming that the average bit depth of a pixel is around 2.5. A raw image coming straight out of Hub Ram with 2.5 bit depth wouldn't be very pretty. the number of available colors is 2^bitdepth... so you are looking at raw images with around 6 available intensity bands, a little better than a binary image. So, algorithms that use binary thresholded image logic should be fine... and there are a lot of applications for those.
I once measured the average bit depth of a rather large series of 8 bit images and found it to be around 6.1.
Rich
Shape measurements are required to confirm these count... but sometimes these can be sampled measurements.
Rich
You said ..."There is enough ram in the prop for a propcam frame I gather."
That seemed like enough of an invitation to me[noparse]:)[/noparse]
I know that you know this stuff... but I thought I would elaborate for others who might be wondering. For those interested in full color panoramics from their super prop bot... it would be a little on the slow side, but the Propcam will be able to acquire all of the data you could want... buy flipping various filters... but the easiest way to composite the images right now would be to download the raw data to your computer and point ImageJ... from NIH... at your data and then run a plug-in... that I promise to write.
Thanks for asking,
Rich
So I was left with a dilemma: start over completely with a different chip, or do a "limited edition" using the current one, while I searched for its replacement. In the end, I felt that the current chip offered enough unique advantages that it was worth plowing ahead with it and that it would fill a niche, even when its replacement was found. Plus, I'd already put so much work into the development that I couldn't bear just to scrap it.
So today, after much hand-wringing, soul-searching, and attempts to locate additional inventory through brokers, I purchased DigiKey's entire stock of 1500 (and I'm old enough to remember when the amount I invested would've bought a house). So now I'm really committed — or, perhaps, should be committed! In the meantime, my efforts to locate additional inventory will continue; but at least there's an assured supply for 1500 PropCAMs, version 1.
-Phil
Graham
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Does this mean PropCAM's almost ready, or has the stock situation just forced your purchase early?
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Cheers,
Simon
BTW: I type as I'm thinking, so please don't take any offense at my writing style
www.norfolkhelicopterclub.co.uk
You'll always have as many take-offs as landings, the trick is to be sure you can take-off again ;-)
I suspect it will be a self-limiting matter, so I see no value in rationing. Those in need of large quantities or a continuous supply will want to evaluate their future demand versus supply before making the plunge. But people have to do that anyway — especially when they're dependent on large manufacturers who can yank a product line with seeming caprice. At least now, with a secure but finite supply, potential users know ahead of time what the risks are. It could've been much worse: I might've made the first batch of 300, then discovered I couldn't get any more.
And it won't end with PropCAM1. As soon as I find a suitable replacement sensor, development will begin on PropCAM2. I read that Micron is reorganizing its imaging division to focus on low-res sensors. This is good news for all of us, and I'm anxious to see what comes of it.
-Phil
From a technical standpoint, the PropCAM has been ready for quite some time. There's an adapter board that Parallax is having made which breaks out the 2x6 2mm header on a daughterboard (e.g. PropCAM) to a 0.1" SIP. Until the MoBoProp motherboard is ready, most PropCAM users will need this adapter. The adapter board (DB-Expander, Parallax p/n 28303) will be available soon, I'm told. There's also an extension cable, which just came online this week here. This will permit separation between the PropCAM module and the DB-Expander.
Now, with all the necessary accessories becoming available, and with lenses and sensors in hand, there's nothing in the way of getting some boards made!
-Phil
Any availability time frame now that you have your sensors on order?
Len
Will your design make it neccessary to have the mobo-board in order to use the Propcam with the propeller products
I've been looking at the new Micron sensors, and they really do have some nice features. But all the ones I've inquired about that have the TrueSNAP global shutter are quite expensive. This is puzzling, since they seem to be aimed at the cost-sensitive automotive industry.
bambino,
The MoBoProp board will not be necessary in order to use the PropCAM. With the DB-Expander and optional extension cable, the PropCAM can be used with virtually any Propeller board. You will want to keep Ports 0-7 available for it, however.
Once the MoBoProp becomes available (no timetable yet), it will likely be the platform of choice for most PropCAM users, due to its small size and the availability of extruded housings for the complete system.
-Phil
They cost 11GBP each, that's probably about $20 probably less.
As an aside, Omnivision do some 5megapixel arrays, one site has them for over 100GBP each, but another wants 9.60GBP if you buy 34 of them. They just expect developers and scientists to pay it.
Graham
They seem to be a good bit more than that over here: $23.71 each in 96-unit lots. But still, a worthy candidate for the PropCAM2 and less money than I was led to believe earlier.
Thanks!
-Phil
Graham
I'll probably use the parallel interface.
Graham
There has, of course, been much discussion here about how to obfuscate Propeller code. Even if I weren't opposed to such measures philosophically — which I am — it would be a risky path to take, given the likelihood that someone would be able to crack the system, exposing the original programmer to liability for damages should the cracker publish his findings. (I haven't actually seen the NDA yet. Perhaps there's a loophole that employs the phrase "all reasonable measures".)
In any event, it will be interesting trying to weave through this legal thicket to come up with a camera based on these chips.
-Phil
Personally I think the whole thing is ridiculous, if a competitor or a potential cloner wanted to get the data sheet they would just get someone to sign the NDA and then disclose the information to them. When they see the clone device available what are they going to do, call every one who signed the NDA into a room and not let them out until someone owns up?
They need to get a grip, they don't do it for their other products. At least Micron's partial data sheets have enough info to get a decent idea about the chips capabilities and the interfacing issues, Omnivision's just say how good they are and that's about it.
Alternatively you add a CPLD or something to produce an interface interface. So silly though I guess it could be ascii based and quite user friendly.
Graham
It's too bad something like a CPLD would be necessary, but it may well be. I guess, as long as there has to be one, it might as well be put to other good uses. For example, while compressing ten bits af A/D output into four, a lot could be done with custom response curves. I had actually contemplated doing something like that with the PropCAM, but with the limited board real estate, it would have entailed putting SMDs on both sides of the board. And that, in turn, would've necessitated a four-layer board just to have an unbroken groundplane behind the lens holder that light can't get through. Such a move would've made the PropCAM much more expensive than I wanted it to be.
Still, it's a shame to offload functions from the Propeller that it could easily handle just to accommodate Micron's paranoia.
-Phil
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Cheers,
Simon
BTW: I type as I'm thinking, so please don't take any offense at my writing style
www.norfolkhelicopterclub.co.uk
You'll always have as many take-offs as landings, the trick is to be sure you can take-off again ;-)
Also there's a small but necessary adapter board that's been held up in production waiting for a connector. This is the DB-Expander that converts the 2mm header on the PropCAM to a SIP so it can be interfaced with the current crop of Propeller boards. So, meanwhile, I've been busy getting other projects out the door. This includes some new stuff for Parallax as well as fishing lure molds for a tackle company I'm associated with.
I know this all seems interminable. What was it? Ten months ago this thread was started? Cripes! That's beyond embarrassing. But progress is being made; and it will get done. It has to: I've got too much time and money (> $15K) invested in it not to!
As always, I do appreciate everyone's patience and forbearance!
-Phil
Graham
No worries Phil; I just know this is gonna be GREAT when I get it
Keep up the good work
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Cheers,
Simon
BTW: I type as I'm thinking, so please don't take any offense at my writing style
www.norfolkhelicopterclub.co.uk
You'll always have as many take-offs as landings, the trick is to be sure you can take-off again ;-)
Good to hear the news about the line scanner... I was just about to think about it again.
What you do... we don't have to. It will be well worth the wait.
Rich
If we all focus our thoughts on those would thwart your efforts, would it help?
Perhaps, like in the movie Scanners? http://www.imdb.com/title/tt0081455/
fish
-Phil