Back
[01:28:52] <[VEGETA]> quick question, lidar sends laser beams in the shape of 3d cone to get 3d point clouds right?
[01:29:23] <[VEGETA]> however, 2d laser range finders sends a beam in one "slice" shape to get x,y only
[01:29:26] <[VEGETA]> right?
[01:38:50] <DagoRed> A lidar is just a fancy and fast set of range finders.
[03:20:01] <z64555> DagoRed: eh, not quite, you'd have to have quite a few of them in an array to match a lidar
[03:21:00] <z64555> or mount them on a turntable, and make a lidar. :)
[04:40:04] <[VEGETA]> DagoRed, what I meant is the whole sensor as one unit
[04:40:19] <[VEGETA]> does it send like a cone shaped beam to get 3d scan?
[04:40:29] <[VEGETA]> either by an array of lasers or not
[04:53:21] <[VEGETA]> z64555 , so lidar is an array of rangefinders or something else?
[05:47:54] <SpeedEvil> LIDAR is a term for many things
[05:48:08] <SpeedEvil> Light ... detection and ranging technically
[05:48:28] <doomlord> will LIDARs cost reduce
[05:48:29] <SpeedEvil> basically you wave a LASER beam around and detect its distance as it's reflected as it's waved
[05:48:34] <SpeedEvil> doomlord: they already are
[05:51:22] <doomlord> i've heard of the cheapo low res close up one
[05:51:32] <doomlord> $300 ish i think for something that works in a few metres?
[05:52:14] <doomlord> i wonder which is easier, better rangefinding or advancements in mahcine vision (i like the idea of machine vision chips & other AI accelerators)
[05:52:21] <doomlord> maybe they'll meet in the middle.
[05:55:31] <SpeedEvil> machine vision can be done with cellphone cameras costing $1
[05:55:55] <SpeedEvil> And can do really decent 3d modeling, with very cheap CPUs
[06:03:45] <veverak> SpeedEvil: any more info on this topic?
[06:03:48] <veverak> ;)
[06:06:49] <[VEGETA]> ok, look here
[06:06:51] <[VEGETA]> http://wiki.ros.org/Sensors
[06:06:56] <[VEGETA]> there are 2d and 3d ones
[06:07:20] <[VEGETA]> my question is what is the difference between them if they are using the same concept
[06:08:52] <[VEGETA]> for example sick 200 vs velodyne lidar
[06:18:27] <doomlord> SpeedEvil machien vision takes more processing power than 3d rendering IMO.
[06:18:30] <[VEGETA]> I cam across LiDAR lite v2 but it is retired.
[06:19:14] <doomlord> the deal with machine vision chips is they have different tradeoffs to 3d graphics chips (more emphasis on low precision eg 8bits, vs F32 or even F16)
[06:19:16] <[VEGETA]> it is very cheap, 115$ only
[06:29:42] <[VEGETA]> I still has the problem of how to mark dynamic obstacles?!
[06:29:50] <[VEGETA]> in a static map
[06:30:24] <[VEGETA]> i.e, appear in the 3d map and let ROS filter them in the 3d point cloud
[07:48:43] <[VEGETA]> anyone have an idea?
[07:48:55] <[VEGETA]> come one i know you do... you just need some tea
[07:57:14] <[VEGETA]> I've researched and found this:
http://wiki.ros.org/sbpl_dynamic_planner
[07:57:50] <[VEGETA]> now how the hell can I integrate it with (octomap + rtab-map) I originally planned to use xD?
[08:24:15] <robopal> hi
[08:44:04] <SpeedEvil> By 'machine vision' I mean the sharply limited subset that is equivalent to LIDAR
[08:44:25] <SpeedEvil> So the much simpler task of taking two cameras spaced 10cm apart, and correlating horizontally.
[08:44:40] <SpeedEvil> Rater than the general machine vision task of trying to work out what shit is
[11:36:52] <z64555> [VEGETA]: A LIDAR uses the same tech as an optical rangefinder, but scans in a cone vs. scanning a point.
[11:37:48] <z64555> A very simple lidar in fact is just an optical range finder that's gimbaled to turn up/down and left/right in a sweeping motion, or make use of a spinning mirror
[11:43:24] <z64555> To handle dynamic object detection, you compare the current frame with the last frame, and compute velocities of each point
[11:43:58] <z64555> image processing techniques apply
[12:51:08] <rue_bed> someone here already did it
[12:51:21] <rue_bed> did 2d scans of their room
[12:51:42] <rue_bed> using an optical measuring tape
[12:52:19] <rue_bed> its raining, I have to get ready for a faire
[13:32:35] <[VEGETA]> z64555 then why the hell they are so pricey!
[13:32:52] <[VEGETA]> even 2d scanners are pricey
[15:35:07] <Loshki> [VEGETA]: Pull the pseudo-lidar out of a Neato XV robot vacuum
[16:19:38] <[VEGETA]> Loshki, ha?
[16:29:25] <[VEGETA]> here is the question:
http://answers.ros.org/question/235472/detect-dynamic-obstacles/
[18:03:41] * rue_house squeels like a girl
[18:03:52] <rue_house> LM3914 ARE LIKE 30C ON ALIEXPRESS!!!!
[18:04:55] <mrdata> tasty
[18:09:16] <rue_house> 3.2 bit flash adc
[18:14:09] <rue_house> so a crude digital scope
[18:14:16] <rue_house> the 3914 is 10 channel
[18:14:21] <rue_house> a 4017 is 10 channel
[18:14:31] <rue_house> so I could go 20 x 40
[18:14:46] <rue_house> 800 leds...
[18:15:19] <rue_house> if I do 20*50 its an even 1000 leds
[18:20:58] <SpeedEvil> Or 32 bit processor with 80MHz 12 bit ADC $10
[18:21:56] <rue_house> yea, but it'd need 70 io lines
[18:22:40] <rue_house> more even
[18:24:55] <SpeedEvil> err - what?
[18:26:53] <rue_house> hmm if I'm gonna make a led scope, I dont think I want to use green leds
[18:27:01] <rue_house> so do I want to use white, blue or red?
[18:28:57] <rue_house> I'm thinking blue
[18:29:11] <rue_house> I'm also thinking white can have any colour mask put over it
[18:35:04] <rue_house> if green says zink sulfate and red says led, what does blue say?
[18:36:15] <Tom_itx> saphire
[18:49:51] <rue_house> k, I'm sold on blue
[18:51:16] <rue_house> ok, I dont know why, but I'm gonna make a 4017/3914 based 25*40 led digital scope
[19:40:30] <Snert> I'd go 3mm leds instead of 5. If you're gonna do round leds.