Post by thread:[EE]: PIC camera tricks

Has anyone tried this camera trick?? I want to connect one of those cheap "camera on a PCB" things, (you know the monochrome ones with standard composite video output that you can buy almost anywhere) to a PIC and use the PIC for image recognition, to detect the position etc of SMD chips on a white surface. My (late night) idea is to get a cheapy camera and change it's one obvious crystal to a slower speed one, a very easy job. So instead of a 15625Hz line rate you get a 1500Hz line rate etc. Obviously going from 25 frames/second to about 2.5 frames/second. BUT then of course the rate is slow enough for a PIC to do processing and work as an effective robotic vision system to detect the position and orientation of the parts shown on the camera. So will it work? I'd love to hear from anyone who's done this trick or similar camera tricks. :o) -Roman -- http://www.piclist.com hint: To leave the PICList spam_OUTpiclist-unsubscribe-requestTakeThisOuTmitvma.mit.edu

There are some other issues when you slow down the clock. In CCD camera's there is a limit to how long you can image before the CCD will saturate. Just slowing down the clock will extend the imaging time. The new CMOS camera's are relatively easy to interface to, in some cases the video is digitized within the camera sensor. w.. Roman Black wrote: {Quote hidden}> Has anyone tried this camera trick?? > > I want to connect one of those cheap "camera > on a PCB" things, (you know the monochrome > ones with standard composite video output > that you can buy almost anywhere) to a PIC and > use the PIC for image recognition, to detect > the position etc of SMD chips on a white > surface. > > My (late night) idea is to get a cheapy > camera and change it's one obvious crystal > to a slower speed one, a very easy job. > So instead of a 15625Hz line rate you get a > 1500Hz line rate etc. Obviously going from > 25 frames/second to about 2.5 frames/second. > > BUT then of course the rate is slow enough > for a PIC to do processing and work as an > effective robotic vision system to detect the > position and orientation of the parts shown > on the camera. > > So will it work? I'd love to hear from anyone > who's done this trick or similar camera tricks. > :o) > -Roman > > -- > http://www.piclist.com hint: To leave the PICList > .....piclist-unsubscribe-requestKILLspam@spam@mitvma.mit.edu -- http://www.piclist.com hint: To leave the PICList piclist-unsubscribe-requestKILLspammitvma.mit.edu

You can get black and white or color CMOS cameras with digital outputs so it can easily be interfaced to a PIC from http://www.electronics123.com/amazon/default.htm and these are almost the same price as typical analog cameras {Quote hidden}> Has anyone tried this camera trick?? > I want to connect one of those cheap "camera > on a PCB" things, (you know the monochrome > ones with standard composite video output > that you can buy almost anywhere) to a PIC and > use the PIC for image recognition, to detect > the position etc of SMD chips on a white > surface. > My (late night) idea is to get a cheapy > camera and change it's one obvious crystal > to a slower speed one, a very easy job. > So instead of a 15625Hz line rate you get a > 1500Hz line rate etc. Obviously going from > 25 frames/second to about 2.5 frames/second. > BUT then of course the rate is slow enough > for a PIC to do processing and work as an > effective robotic vision system to detect the > position and orientation of the parts shown > on the camera. > So will it work? I'd love to hear from anyone > who's done this trick or similar camera tricks. > :o) > -Roman -- http://www.piclist.com hint: To leave the PICList .....piclist-unsubscribe-requestKILLspam.....mitvma.mit.edu

Potentially, yes. However most CCD readout amps are AC coupled so you might not have any charge to read since the settling time will also be extended 10 fold. This will also increase your camera sensitivity 10 fold (IF the readout bucket brigade can handle the slower clock rate. Think DRAM refresh capacitances). One trick that was used decades ago for slow scan television (amateur radio TV in a 3khz BW) is a sweep fast sample and hold. You basically use a ramp generator at the horizontal rate and monostable S&H driven by a comparator whose second input you control. With a low voltage, you sample the left edge of the video, and with a higher voltage, the right side. You can use a slow ramp generator (15seconds was what was used for SSTV) for automatic sweeping. This gives you 15k samples a second (easily within a PIC's capability) and allows you to use ANY standard video signal as input (as long as it doesn't change over the sampling time). It's easy enough to test the idea. Just look at the 10x slower video output to see if it 'looks right'. Robert Roman Black wrote: {Quote hidden}> > Has anyone tried this camera trick?? > > I want to connect one of those cheap "camera > on a PCB" things, (you know the monochrome > ones with standard composite video output > that you can buy almost anywhere) to a PIC and > use the PIC for image recognition, to detect > the position etc of SMD chips on a white > surface. > > My (late night) idea is to get a cheapy > camera and change it's one obvious crystal > to a slower speed one, a very easy job. > So instead of a 15625Hz line rate you get a > 1500Hz line rate etc. Obviously going from > 25 frames/second to about 2.5 frames/second. > > BUT then of course the rate is slow enough > for a PIC to do processing and work as an > effective robotic vision system to detect the > position and orientation of the parts shown > on the camera. > > So will it work? I'd love to hear from anyone > who's done this trick or similar camera tricks. > :o) > -Roman > > -- > http://www.piclist.com hint: To leave the PICList > EraseMEpiclist-unsubscribe-requestspam_OUTTakeThisOuTmitvma.mit.edu -- http://www.piclist.com hint: To leave the PICList piclist-unsubscribe-requestspam_OUTmitvma.mit.edu

Nice chips. However www.electronics123.com/amazon/datasheet/OV5017DS.pdf "Since the video data is a continuous non-stop byte stream, ..." you still need to be able to capture the data a video rates since there does not appear to be a way to slow the transfer rate since there is no on board frame buffer, and the frame rate control merely controls the update period. i.e. how many frames get dropped before a new one sampled. I suppose one could use a counter chain to control which sample you get after HREF, but CMOS sensors are also fairly noisy compared to CCD, so I don't see that you're much futher ahead with these chips. Robert Andrew Mcdowell wrote: {Quote hidden}> > You can get black and white or color CMOS cameras with digital outputs so it > can easily be interfaced to a PIC from > http://www.electronics123.com/amazon/default.htm and these are almost the > same price as typical analog cameras > > > Has anyone tried this camera trick?? > > > I want to connect one of those cheap "camera > > on a PCB" things, (you know the monochrome > > ones with standard composite video output > > that you can buy almost anywhere) to a PIC and > > use the PIC for image recognition, to detect > > the position etc of SMD chips on a white > > surface. > > > My (late night) idea is to get a cheapy > > camera and change it's one obvious crystal > > to a slower speed one, a very easy job. > > So instead of a 15625Hz line rate you get a > > 1500Hz line rate etc. Obviously going from > > 25 frames/second to about 2.5 frames/second. > > > BUT then of course the rate is slow enough > > for a PIC to do processing and work as an > > effective robotic vision system to detect the > > position and orientation of the parts shown > > on the camera. > > > So will it work? I'd love to hear from anyone > > who's done this trick or similar camera tricks. > > :o) > > -Roman > > -- > http://www.piclist.com hint: To leave the PICList > @spam@piclist-unsubscribe-requestKILLspammitvma.mit.edu -- http://www.piclist.com hint: To leave the PICList KILLspampiclist-unsubscribe-requestKILLspammitvma.mit.edu

Andrew Mcdowell wrote: > > You can get black and white or color CMOS cameras with digital outputs so it > can easily be interfaced to a PIC from > http://www.electronics123.com/amazon/default.htm and these are almost the > same price as typical analog cameras Thanks Andrew for the link, they have quite a few cameras there worth investigating. :o) -Roman -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

Thanks for the info Walter, I was worried there would be some issue like that. But if this only applies to CCD cameras maybe I could just use a CMOS one? -Roman Walter Banks wrote: {Quote hidden}> > There are some other issues when you slow down the clock. In CCD > camera's > there is a limit to how long you can image before the CCD will saturate. > Just > slowing down the clock will extend the imaging time. The new CMOS > camera's > are relatively easy to interface to, in some cases the video is > digitized within the > camera sensor. > Roman Black wrote: > > > Has anyone tried this camera trick?? > > > > I want to connect one of those cheap "camera > > on a PCB" things, (you know the monochrome > > ones with standard composite video output > > that you can buy almost anywhere) to a PIC and > > use the PIC for image recognition, to detect > > the position etc of SMD chips on a white > > surface. > > > > My (late night) idea is to get a cheapy > > camera and change it's one obvious crystal > > to a slower speed one, -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

Hi Roman, make an search for gameboy cameras :) ( this artificial retina chip is used in it ) for example: homepages.paradise.net.nz/~vkemp/gbcam.htm or www.geocities.com/SiliconValley/Park/1302/gbcam.html or http://www.lh.co.nz/hardware/ar/techinfo.htm You should be able to get hold of one quite easily. The really neat stuff is the edge enhancement / extraction and goes right along with your objectives. /Tony -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

I have done 'camera tricks' but I used a PIC only once (to interface the PC that was doing the tricks with the camera controls). I think that something very crude can be done but by the time you want to determine part orientation you start to need MIPS and room. If the part orientation is not a problem (already aligned) then it might work. Slowing down the camera crystal also increases exposure 10 times and leakage by the same amount (assuming that the dynamic pulse circuitry and S&H systems and clamping in the camera do not c..p out totally due to too slow speed). The latter causes image shading or even 'disappearance' which is something you do NOT want for this application imho. Imho leave the clock alone and perform slicing and integration using analog means at normal video speed, then use the PIC to control the servos for positioning. Of course a PC has the power to do this by itself. good luck, Peter -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

Hi, > I have done 'camera tricks' but I used a PIC only once (to interface the > PC that was doing the tricks with the camera controls). > > I think that something very crude can be done but by the time you want to > determine part orientation you start to need MIPS and room. If the part > orientation is not a problem (already aligned) then it might work. Do you have references or source code for the position and orientation determination ?? I have been looking at it since Roman gave that idea but have not been able to find much. Seem like everyone is trying to do face recognition nowadays and are not working with rectangles anymore :-( Best regards, Alexandre Guimaraes -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

Thanks!! I hadn't even though of the good old gameboy camera. :o) -Roman Kübek Tony wrote: {Quote hidden}> > Hi Roman, > make an search for gameboy cameras :) > ( this artificial retina chip is used in it ) > > for example: > > homepages.paradise.net.nz/~vkemp/gbcam.htm > or > www.geocities.com/SiliconValley/Park/1302/gbcam.html > or > http://www.lh.co.nz/hardware/ar/techinfo.htm > > You should be able to get hold of one quite easily. > The really neat stuff is the edge enhancement / extraction > and goes right along with your objectives. -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

Peter L. Peres wrote: > > I have done 'camera tricks' but I used a PIC only once (to interface the > PC that was doing the tricks with the camera controls). > > I think that something very crude can be done but by the time you want to > determine part orientation you start to need MIPS and room. If the part > orientation is not a problem (already aligned) then it might work. MIPS? to detect the angular orientation of a 20 pixel sized rectangle to about 2% accuracy? Thats about 12 PIC instructions... ;o) Also, sorry if I gave the impression that I wanted to do the entire image processing with the PIC, actually it just needs to be a pre-processor and give some sort of meaningful data to the computer. If I get a slow frame rate, about 5 frames per second or less the PC will be able to do all the processing in between driving the machine's stepper motors. It may not even need the PIC. :o) > > Slowing down the camera crystal also increases exposure 10 times and > leakage by the same amount (assuming that the dynamic pulse circuitry and > S&H systems and clamping in the camera do not c..p out totally due to too > slow speed). The latter causes image shading or even 'disappearance' > which is something you do NOT want for this application imho. True. Maybe CMOS cameras would not suffer from this problem? > Imho leave the clock alone and perform slicing and integration using > analog means at normal video speed, then use the PIC to control the servos > for positioning. Yes, a good idea to slice it. In this app the PC can spend a time period getting the scan and finding parts posiions, then a few parts can be placed before the next scan is needed. > good luck, Thanks! :o) -Roman -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

Alexandre Guimarães wrote: > > Hi, > > > I have done 'camera tricks' but I used a PIC only once (to interface the > > PC that was doing the tricks with the camera controls). > > > > I think that something very crude can be done but by the time you want to > > determine part orientation you start to need MIPS and room. If the part > > orientation is not a problem (already aligned) then it might work. > > Do you have references or source code for the position and orientation > determination ?? I have been looking at it since Roman gave that idea but > have not been able to find much. Seem like everyone is trying to do face > recognition nowadays and are not working with rectangles anymore :-( Relax with the angular orientation stuff! ;o) It's very easy, as is finding the center of mass. The rectangle (SMD part on camera) has 90 degree sides, and only one side needs to be analysed to give orientation. With a 0805 etc sized SMD part, orientation is not critical and 64 rotary positions are acceptable. So you only need to be able to identify 64/2 = 32 actual orientations, probably by a simple lookup table. For centre of mass you average the positions of all the highest pixels, and lowest pixels, then average them together. A few additions and subtractions. Besides being a TV technician I did a couple years as a 2d game graphics programmer. :o) -Roman -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

Alexandre, I am sorry but I cannot help you with source for the camera aided positioning. You will have to do your own homework, although there are some projects related to computer vision for CNC and robotics on the web (mostly academic software for unix and I don't have a url). You really need to read about robot vision and a good start would be a book, for example Image Processing and Computer Vision by Parker (I keep recommending this because I used it - there are others). Peter -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

Alan King wrote: > > Hi Roman, > We're going to have to get talking, you seem to be working on a lot of > similar things to what I'm working on at the moment. Talking we are. :o) > Here's what I'm working on at the moment: > http://photos.yahoo.com/alantak69 and look in the stepper folder. I > have a few hundred steppers around now and a couple thousand mosfets, > I'm going to start selling stepper motor controllers and machines I'll > make from them. A few hundred motors? Wow. I saw your PCB photo. {Quote hidden}> I too am planning a simple surface mounting machine, I have thousands > of surface mount parts and controllers to use so it's one of the next > things on my agenda after the driller. I also have probably 20 or 30 > printer mechanics laying around, one of those hooked to my controller > was my plan for pasting and placing as well. Heck I want to get another > laser printer and take it apart too, and make it print straight onto > copper board if possible so I can skip the iron on step. There has to > be a way, even if it can't go straight to the copper there must be some > simple preperation that will make it work and be better than ironing, > although that's producing very good boards for me for now. Cool. Sounds like you will be making stepper controllers? Can you give any specs? One thing with the boards would be to use your CNC drilling mill to hold a DALO pen, then draw direct to PCB and etch. Drill last. :o) -Roman -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

>The rectangle (SMD part on camera) has 90 degree >sides, and only one side needs to be analysed >to give orientation. With a 0805 etc sized SMD >part, orientation is not critical and 64 rotary >positions are acceptable. So you only need to >be able to identify 64/2 = 32 actual orientations, >probably by a simple lookup table. How do you plan to deal with polarity sensitive components like diodes and electrolytic caps? Maybe all your diodes come in SOT23 and you don't use electros ;) I take it you plan to be able to identify which side has one lead and which has 2 leads on an SOT23 - yeah I know one thing at a time get the basic system working first ;) -- http://www.piclist.com#nomail Going offline? Don't AutoReply us! email RemoveMElistservTakeThisOuTmitvma.mit.edu with SET PICList DIGEST in the body