Accurate Infrared Object Detection.
Steve in NM
Posts: 54
I'll apologize to the moderators now if this isn't where this belongs. I figured since it's aimed at folks that participate in sumo, they probably know the answers instinctively.
·
I've been reading everything I can find on IR detection for a couple of weeks now. There's a lot of great information about what sensors are·popular for object detection and ranging. What almost never seems to get mentioned is which IRLED folks are using. From what I can determine, The IRLED and how it's being driven is *AS* important (if not more so) than the detector. Note that I'm not referring to the Sharp integrated rangefinders. I'm talking about the 3-pin type detectors from Panasonic and Vishay which are intended for use with a separate, modulated IRLED. The Sharp IS471F is a neat device and does the modulation itself, but nobody seems to get any range out of it. Here again they always neglect to say which IRLED they're using. I'm also wondering if they'd interfere with each other if used in an array.
·
So...
·
There's an impressive assortment of IR emitters out there. The Vishay TSAL6100 has a 10 degree half-angle. Is that as narrow as they come? It has a "Surge Forward Current" rating of 1.5 amps (!), if it were visible, it might put yer eye out! A random sampling of half-angles is 10, 17, 20, and 30 degrees.
·
Let’s assume I want to detect and track either a golf or ping-pong ball that is from 0” to 36" away. Determining the distance isn’t important since I already know it’s less than 36”. What I want to do is determine, with reasonable accuracy, the relative vector to the target.
In my mind's eye (cheaper that way):
·
1) I have an array of detectors on a robot·–·one each left, right, and rear, and an as yet undetermined number in the front. Call this the “main” array. The detectors are either Panasonic PNA4602M or Vishay TSOP1138. I'm thinking it really doesn't matter which since they are so similar. True?
·
2) I have an array of emitters - one each left, right, and rear, and an as yet undetermined number in the front, alternating in arrangement between wide angle and narrow angle. By watching the state of the detectors and controlling the individual firing of the emitters, I should be able to narrow in on the target with fairly reasonable accuracy. Yes?
·
The *real* question for you folks is how would you do this? I can see (again – in my mind’s eye) how to determine where the ball is generally (left, right, front, rear) by generating one big IR pulse (with wide angle LEDs) and then capturing the state of the left, right, front, and rear detectors. Now I rotate the “main” emitter/detector array to that vector and start narrowing in. This is where I’m drawing a blank. Maybe there are too many possibilities and I’m just guessing at which scenario might work best.
·
Would you use 2 wides and 3 narrows? 3 wides and 2 narrows? ·Scan with the wides, then with the narrows?· Vice-versa? Scan from the outsides in? Why and or why not? Should I save my money and just buy a pet rock?
·
Any input, comments, or criticisms gratefully accepted.
Steve in NM
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
I'm not a complete idiot! Some of my parts are missing.
·
I've been reading everything I can find on IR detection for a couple of weeks now. There's a lot of great information about what sensors are·popular for object detection and ranging. What almost never seems to get mentioned is which IRLED folks are using. From what I can determine, The IRLED and how it's being driven is *AS* important (if not more so) than the detector. Note that I'm not referring to the Sharp integrated rangefinders. I'm talking about the 3-pin type detectors from Panasonic and Vishay which are intended for use with a separate, modulated IRLED. The Sharp IS471F is a neat device and does the modulation itself, but nobody seems to get any range out of it. Here again they always neglect to say which IRLED they're using. I'm also wondering if they'd interfere with each other if used in an array.
·
So...
·
There's an impressive assortment of IR emitters out there. The Vishay TSAL6100 has a 10 degree half-angle. Is that as narrow as they come? It has a "Surge Forward Current" rating of 1.5 amps (!), if it were visible, it might put yer eye out! A random sampling of half-angles is 10, 17, 20, and 30 degrees.
·
Let’s assume I want to detect and track either a golf or ping-pong ball that is from 0” to 36" away. Determining the distance isn’t important since I already know it’s less than 36”. What I want to do is determine, with reasonable accuracy, the relative vector to the target.
In my mind's eye (cheaper that way):
·
1) I have an array of detectors on a robot·–·one each left, right, and rear, and an as yet undetermined number in the front. Call this the “main” array. The detectors are either Panasonic PNA4602M or Vishay TSOP1138. I'm thinking it really doesn't matter which since they are so similar. True?
·
2) I have an array of emitters - one each left, right, and rear, and an as yet undetermined number in the front, alternating in arrangement between wide angle and narrow angle. By watching the state of the detectors and controlling the individual firing of the emitters, I should be able to narrow in on the target with fairly reasonable accuracy. Yes?
·
The *real* question for you folks is how would you do this? I can see (again – in my mind’s eye) how to determine where the ball is generally (left, right, front, rear) by generating one big IR pulse (with wide angle LEDs) and then capturing the state of the left, right, front, and rear detectors. Now I rotate the “main” emitter/detector array to that vector and start narrowing in. This is where I’m drawing a blank. Maybe there are too many possibilities and I’m just guessing at which scenario might work best.
·
Would you use 2 wides and 3 narrows? 3 wides and 2 narrows? ·Scan with the wides, then with the narrows?· Vice-versa? Scan from the outsides in? Why and or why not? Should I save my money and just buy a pet rock?
·
Any input, comments, or criticisms gratefully accepted.
Steve in NM
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
I'm not a complete idiot! Some of my parts are missing.
Comments
Seriously, I think you're jumping in before you've learned how to swim. You're trying to conceive of a complex mechanism to do a tracking job that doctoral theses and big consulting contracts are written on. If you had an isolated ping pong ball or golf ball painted to provide a good (not ideal) IR reflector moving slowly (swinging) in an environment with no other objects, I'd say "go for it". As soon as you introduce walls and floors and maybe other objects, the light scatter will make it all useless. Have you looked at using something like a CMUcam or its successors? An IR emitter / detector array is not a bad idea, but probably much too complex for what it's capable of doing. What have other people successfully done? What have they learned?
If the only thing int the target area is the ping pong ball, you could use two sensors like the Sharp GP2D15 or one of the several variants.
Mount two sensors on the robot on servos several inches apart. When the sensors sweep the area and both detect the object you can determine the distance and direction based on the servo positions.
Daryl