go_away

Author Topic: camera models for tracking  (Read 3489 times)

0 Members and 1 Guest are viewing this topic.

Offline nitriumTopic starter

  • Beginner
  • *
  • Posts: 3
  • Helpful? 0
camera models for tracking
« on: October 24, 2007, 01:15:40 AM »
Hello,

what cameras do you offer for tracking robots. I need your opinions for wide range of camera models including cost features, capability features.

Thanks.

Offline bulkhead

  • Full Member
  • ***
  • Posts: 102
  • Helpful? 0
Re: camera models for tracking
« Reply #1 on: October 24, 2007, 01:32:27 AM »
Any camera will work, you just need software.  Try www.roborealm.com (if you have a webcam hooked up, you can download their software and have it analyzing colors in a matter of minutes.  This would work if you have a computer onboard or somehow stream the video to a computer and then send back the information wirelessly (rf, bluetooth).

For better results, a camera with autofocus (like a digital camera) could be used.  In addition to providing a better quality image, it could actually be used to measure distance by measuring the changes made in order to 'focus' on an object at a certain distance.

Offline Steve Joblin

  • Supreme Robot
  • *****
  • Posts: 405
  • Helpful? 2
Re: camera models for tracking
« Reply #2 on: October 24, 2007, 08:25:46 AM »
I too would suggest roborealm as the software is free and it works with any basic webcamera.  Also give a check out to the CMUcam (http://www.cs.cmu.edu/~cmucam/) and AVRcam (http://www.jrobot.net/Projects/AVRcam.html) products. 

Offline nitriumTopic starter

  • Beginner
  • *
  • Posts: 3
  • Helpful? 0
Re: camera models for tracking
« Reply #3 on: November 08, 2007, 12:47:30 PM »
Thank you both for your offers about cameras. I have evaluated both however, the resolution is too low for extracting human or just meaningful data from the stream.

I am planning to do a robot which finds a human in the stream, extracts the human and follow.

Any further recommendations?


Offline bukowski

  • Robot Overlord
  • ****
  • Posts: 219
  • Helpful? 0
Re: camera models for tracking
« Reply #4 on: November 08, 2007, 01:21:55 PM »
Get a cozy chair, coz your going to be writing code for the rest of your life.

Offline Steve Joblin

  • Supreme Robot
  • *****
  • Posts: 405
  • Helpful? 2
Re: camera models for tracking
« Reply #5 on: November 08, 2007, 01:25:54 PM »
what are you trying to do?  build a rescue-bot that can detect a human in a river and go to him?

Offline nitriumTopic starter

  • Beginner
  • *
  • Posts: 3
  • Helpful? 0
Re: camera models for tracking
« Reply #6 on: November 08, 2007, 01:28:10 PM »
A robot which will follow a person in the airport by carrying luggage.   :P

Offline sdk32285

  • Supreme Robot
  • *****
  • Posts: 418
  • Helpful? 0
Re: camera models for tracking
« Reply #7 on: November 08, 2007, 03:05:21 PM »
Hi
You might want to look at the Lucas Kanade Tracker. There are some papers by Simon Baker and Iain Matthews that talk about update strategies so that you can handle when the original template changes dramatically in scale/rotation and rotation.
Robots for Roboticists Blog - http://robotsforroboticists.com/

Offline paulstreats

  • Supreme Robot
  • *****
  • Posts: 1,381
  • Helpful? 21
Re: camera models for tracking
« Reply #8 on: November 08, 2007, 07:04:52 PM »
If its as simple as following a predefined target like the airport scenario, you might want to consider some kind of infrared beacon on the target subject for the suitcase to follow. Also there are ways to modulate the infrared to increase accuracy, allowing probably more than one device and person to operate. also you can attempt to polarise the infrared light to only emit and detect light travelling in a polarised form

Offline Robotboy86

  • Robot Overlord
  • ****
  • Posts: 139
  • Helpful? 0
Re: camera models for tracking
« Reply #9 on: November 09, 2007, 12:13:25 PM »
IMO, your trying to do something that masters of vision cannot always do right.  Humans use vision there entire lifes.. and yet we can have problems analyzing someone and picking them from a crowd.

Offline Killtacularmods

  • Jr. Member
  • **
  • Posts: 28
  • Helpful? 0
Re: camera models for tracking
« Reply #10 on: November 09, 2007, 06:43:47 PM »
Hey, i got a few questions, i want to use the AVRcam with my microcontroller im making, but it looks like the AVRcam has to be attached to "there" microcontroller by a 30pin header on the back of the camera? do you have to use there controller or can it be connected to my board through the urat pins some how? or can there controller be connected to my controller? Id really like to attach a camera directly to my microcontroller without haveing another microcontroller board adding clutter to my robot.

Offline Ro-Bot-X

  • Contest Winner
  • Supreme Robot
  • ****
  • Posts: 1,431
  • Helpful? 25
  • Store: RoBotXDesigns.ca
Re: camera models for tracking
« Reply #11 on: November 11, 2007, 05:39:10 AM »
AVRcam has it's own mega8 to deal with image processing. This will connect to your PC using an onboard Max232 level converter, or to your mega8 using the direct Tx to Rx and Rx to Tx pins. By letting the AVRcam have it's own MCU it frees up space and time on your main MCU to do other tasks, such field orientation, obstacle avoidance and so on. Consider it a separate sensor. Like some Ultrasonic sensors have their own MCU onboard for operating. It makes your life a lot easier.
Check out the uBotino robot controller!

Offline dunk

  • Expert Roboticist
  • Supreme Robot
  • *****
  • Posts: 1,086
  • Helpful? 21
Re: camera models for tracking
« Reply #12 on: November 11, 2007, 08:24:05 AM »
Quote
do you have to use there controller or can it be connected to my board through the urat pins some how?

if you look at the http://www.jrobot.net/Projects/AVRcam.html website they tell you exactly how it is made. there are full circuit diagrams and source code.
so you could in theory connect a camera to an AVR in the same way using the AVRcam instructions for reference but including your own features.

i think the AVRcam documentation demonstrates what a difficult undertaking this would be though.
the AVRcam is also right up against the limit of it's microcontroller's resources so adding any extra features would be difficult.

it would certainly be far easier to just use an AVRcam and pass images to a 2nd AVR over the UART.

dunk.

 


Get Your Ad Here