Re: Barcodes & error detection
"Brian Davis" <[email protected]>
| Newsgroups | gmane.comp.hardware.lego.robotics |
|---|---|
| Organization | None |
| Message-ID | <[email protected]> |
In lugnet.robotics, [email protected] wrote: > Are you lighting the area with a light brick, or > relying on ambient illumination? Gus uses two ight sensors, one to follow the (left edge) of a black line on a white background, and a seperate light sensor on an "outrigger" to read the barcodes. There is no light brick, but instead both use the light sensor itself to supply light within a well-shielded enclosure. > try to always shield the area from ambient and > provide my own illumination. I've not yet supplied my own light in the form of a seperately powered lightbrick - the illumination LED of the LEGO light sensor has always seemed sufficient. For the line-following light sensor, the shroud is black enclosure with a front opening ("windows") two studs wide by one brick high. If the window is masked down any more (in height by using plates, or in width by using 1x2 plates with rails), the FOV for the line following sensor is too small and there's hardly any "grey" readings to use for small corrections. The outrigger barcode-reading light sensor uses an identical shroud but one that is masked down still more by using a plate and a tile so it is 2 studs wide but only 1 plate high. This much narrower slit allows a rapid transition as the strips of the barcode pass under (the transition region, in which the light sensor could "see" both strips at once and return an indeterminate reading, is reduced). Make that window wider (well, taller, in the direction of the forklifts motion), and the transitions take longer, requiring either wider strips (ugly, and awkward, as the slight wobble of the robot may make things more difficult... and there will be more wobbles or variations the longer it take to cross the entire barcode) or better recognition of the "pure" strip zones where the view is uncontaminated by the transition regions (again, tough; as it is (I'm using the standard firmware) I only have time to execute about three to five reads of the "pure" strip). The shrouds are only 4 mm from the table surface. You can hit it with a flashlight to no effect, although it does have trouble when moving from a "ambient" light area into direct sunlight (it will, actually, still follow the line, but the barcode reader has more trouble). The light sensor itself is two studs up from the end of the shrouds, putting the phototransistor about 2 cm above the table surface. -- Brian Davis