DSP board for video stopwatch

[oznerol]

I imagine this is the correct forum to post in. So, here's the situation.

My team and I have been assigned a project for a video stopwatch. Simply put, it is not much more than a regular stopwatch which utilizes video input to increase its accuracy.

A big problem we are facing is not the processing algorithms, but rather the hardware we are going to run it on.

I'm looking for some suggestions from anyone with experience with this stuff, in particular some cheap development kits we could possibly attain. We have a pretty vague idea of the hardware we can use, so any help would be greatly appreciated.

Basically we need the following: video input from a camera - we were considering using a web cam and get a dev kit with a usb interface, though getting a small enough camera is essential. Some kind of DSP board, either digital or analog input. Low power consumption, all packaged together in a relatively small unit.

If you guys could point us in the right direction, we'd be interested in hearing it. Or, if you are experienced with DSP boards and need me to elaborate more, drop me a PM.

Thanks in advance. :beer:

 

[PottedMeat]

i dont know much about DSPs but Analog Devices may be running a deal for one of their TigerSHARC demo boards, last time i looked at one it had a mess of A/V inputs/outputs and came with an eval version of their VisualDSP. I think the deal was ~$99, it may be more now.

 

[dowxp]

try a MCU like the generic HC12 or try programming an FGPGA ; I haven't taken DSP (never will), but thats what my classmates are using for their project. AFAIK, labview has a capture driver for something like a webcam.

 

[blahblah99]

Analog Devices has a good range of DSPs you can play with... So does TI. Motorola (Freescale) also has some DSPs, but not as powerful as analog devices/TI.

 

[bobsmith1492]

Doesn't TV have some sort of carrier frequency at 32.768KHz? Just put a bandpass filter and a comparator at the input to detect this change, use a Schmidt trigger to pick up the transitions, and run it into a microcontroller to divide by 32,768 (2^15 - one run of a 16-bit counter register) and there you have it - a 1 second timer.

Basically, why are you using a DSP board? Try to design the thing yourself; it doesn't sound too complicated....

EDIT: Ok, one more thing; what exactly is the video supposed to do again? I may have it confused with something else...

 

[jmcoreymv]

Check out the Analog Devices BF537 Eval Kit. It has a Blackfin DSP chip and a A/V in and out and comes with sample code for using it. If youre a student I think you can get it for like 300-350 bucks.

 

[oznerol]

Originally posted by: bobsmith1492
Doesn't TV have some sort of carrier frequency at 32.768KHz? Just put a bandpass filter and a comparator at the input to detect this change, use a Schmidt trigger to pick up the transitions, and run it into a microcontroller to divide by 32,768 (2^15 - one run of a 16-bit counter register) and there you have it - a 1 second timer.

Basically, why are you using a DSP board? Try to design the thing yourself; it doesn't sound too complicated....

EDIT: Ok, one more thing; what exactly is the video supposed to do again? I may have it confused with something else...

I think we are on 2 different pages here.

The stopwatch is a basic stopwatch, with a start/stop button, etc. However, a small camera will be mounted on the front (this is for the game of curling, so the entire device is mounted on a broom).

The user will use the stopwatch as usual, however pointing the broom at the target for a short period. The camera will take in a video feed, and with some interpolation and image processing, the watch will determine the precise time the object crosses both lines - basically improving the accuracy of the standard stopwatch by eliminating human error.

Ultimately our goal is to have the entire process automated, so no start/stop needs to be pressed the unit would just need to be on, set, and pointed at the target.

 

[oznerol]

Originally posted by: jmcoreymv
Check out the Analog Devices BF537 Eval Kit. It has a Blackfin DSP chip and a A/V in and out and comes with sample code for using it. If youre a student I think you can get it for like 300-350 bucks.

We looked at the Analog Devices kits, but honestly we have no experience with this stuff so it's hard to know what to look for. Is this what you are talking about?

I actually have a question about these boards in general. They all have PC requirements, I imagine this is for programming and working on the board, so this is what the USB port is for, correct? Also, these things can work as standalone devices, right?

Here was my idea, maybe one of you can point out what is wrong with it.

5V CCD Camera
BF533 Development Board

The board has RCA inputs, the camera is RCA out, we can do the A/D conversion on board. Am I missing something?

 

[Megamixman]

The board you linked to is a Blackfin 537. Honestly, the best way to do this would be to create the algorithm, emulate it to see what kind of performance you would need, and then choose your DSP. The 537 is a very powerful DSP. I'm not sure if you would need that much power. There is also the possibility that your algorithm requires more power, since it seems you are trying to detect the time of an object at two points.

Edit: The Blackfin 537 does have an ADC, but the 533 doesn't. Also is your algorithm going to be mostly integer based or will it use a lot floating point math? Do you want to use a MCU to control the data flow and the DSP just for the algorithm? There is more to choosing the right DSP then just performance. If your application is battery powered, power usage is another consideration.

 

[bobsmith1492]

Ah ha... that makes sense now. What about using an infrared transmitter/receiver pair that outputs a specific frequency, and watch for that frequency at the input? Basically, use a motion detector like they use in automatic sinks to initialize or stop the watch? I'm still not seeing the need for a camera and complicated algorithms to detect the proximity of an object.

 

[oznerol]

Originally posted by: bobsmith1492
Ah ha... that makes sense now. What about using an infrared transmitter/receiver pair that outputs a specific frequency, and watch for that frequency at the input? Basically, use a motion detector like they use in automatic sinks to initialize or stop the watch? I'm still not seeing the need for a camera and complicated algorithms to detect the proximity of an object.

Well, it is meant to be a portable device for the game of curling, not something we would mount. Think of it as a pda-sized device, with mounted camera, that you could just point at the stone as you walked alongside it, and calculate the time between lines.

We also could have used a laser-based system, kind of like a radar gun, but we are more comfortable with the video processing.

We spoke to our professor today and he suggested that we worried about the format of the camera output as well as the algorithm process to know what kind of board we'd need.

With that in mind, I found this and I think it might be useful for the project. The only problem is I'm not sure how to handle the output signal. Anyone have experience with CMOS cameras?