Depends how many directions you want to monitor at once. I believe this method could also work with a fisheye lens on the front of the robot as well.
(or better yet, one of the 360 degree mirrors, and the camera in the middle of the bot… but those mirrors are expensive!)
I imagine this wouldn’t be used for general purpose obstacle avoidance. But it might fit certain applications, like being used out the side window of an autonomous car.
I am not saying its a great idea
And Chunga: That doesn’t quite fit his criteria of ‘doing something not done yet’. I know laser measuring tape devices can do that trig. And you don’t want to do that with sonar anyway, at least not without narrowing the sonar beam very tightly.
If you really want to get into something this involved, google 3D SLAM algorithoms. It’s 360 deg laser mapping. You can do alot with the information after scanning.
It think integrating video image manipulation and range sensors into a mutually beneficial relationship would be a better software start though.
Just a thought, why not integrate two lasers which point under a certain angle towards the point of view of the camera.
The distance between the two dots can be used as an indication of range, but then you’d have errors if the two dots fall on a surface that’s not straight in front of you… hmm…
But if you’d be able to alter the angle and adjust the angle until the point where the two points meet, then you’d have a correct reading.
Am new to this stuff so forgive me if this is wrong, but a slightly different idea from the previous posts.
If you used something like 2 CMUcams on pan and tilt then they can be set up so they point automatically at the object being tracked. Then it would be a matter of reading the 4 servo settings, take into account the distance between the two mountings and from that you could calculate not only the distance but also the vertical and horizontal angle to the object. Not sure how acurate it would be, but should give a rough idea ?
Its easier if you mount both cameras on a single mount, with one tilt servo. Then each camera can have its own “pan” servo, and you know the are both always looking in the same horizontal plane. Then it is just a matter of looking at the angles formed by the two servos, rather than four.
I have a setup like that documented on my MicroRaptor vision page:
In fact, with the way I’ve got it set up, you only need a single servo, which converges and diverges the cameras. The whole assembly is mounted on a pan & tilt mount…
Wouldn’t like using only 1 pan and 1 tilt will facilitate de job, using the tilt rarely, mostly the pan.
Let’s say I want to do this for a robot so he can avoid abstacles, it’s his only sensors, 2 cams. If possible to use it to do some (if possible) mapping, not very presice, but of waht the bot sees.
Oh, and evo, I didn’t find anything that explained 3D SLAM algorithms
Checking google to see what SLAM is, you might be able to make a simple distance measuring/mapping device using one cam, two servos (one slightly modified), a laser pointer, and the usual collection of thin wood, thumb tacks, hot glue, paper clips, and using roborealm to track the laser dot.
Doing more thinking, I’d skip the cam and maybe use a modulated laser beam and a demodulatng detector (possibly made from a modulated IR emitter/detector pair). Have the laser and detector seperated by a specific distance. Make a narrow view slit for the detector and then have it slowly panned via servo until it sees the reflected modulated laser dot thru the slit. Capture the servo position at the time of detection, and use that position to calculate the distance to the dot. Have this mounted on another servo so the emitter/detector can be rotated slightly and another distance scan performed. I save the duct tape for the really big stuff!
I’m talking about operating the laser diode like the IR diode in remote controls. Flash it at 38khz and use a remote control IR detector operating at 38khz to detect the laser dot. This would help ensure that only the laser dot is being detected instead of other random visible light sources.
Ok, so I could just build my own circuit using the IR detector and a couple of infrared LEDs? Use like a NAND gate circuit to make the ocillation at 38Khz. But, would the range be very limited? Like maximum 30 cm?
Are there LED that can project themselves farther?
“Are there LED that can project themselves farther?”
Yes, they are laser diodes. Below are some examples of laser gizmos. Notice that a very cheap laser pointer is used. I’ve never done any tinkering with these, but would make an interesting school project.
You probably need to fully map out what you plan to do first. The main issue is that with the cheap laser pointers you are dealing with visible light and not IR. Cheap IR detectors that are made to respond to 38khz modulated IR may not see the visible light from the laser. The detector below might be hacked to remove the IR filtering so it will respond to the 38khz visible light. For testing, the laser beam could be modulated at 38khz using an IR remote control to modulate the power to the laser. I don’t know what type of laser you have, so I can’t comment on how you need to modify it. For the el-cheapo types, remove the batterys, insert a small dowel with a + power wire on the end, and the housing is the - power connection. Use a tie wrap to hold the button down.
