Vision Processing: Has anyone seen what the scale lighting looks at angles

Technical Programming
#1

Hi Everyone,

Bare with me this is a confusing question. Our team was trying to think our way through being able to track the edges of the scale with vision

the lights look linear enough where it would be doable, however looking at the lights I am noticing a “lip” on the edges with the lights in between, that I am worried makes it difficult to see from lower angles

Does anyone have any pictures from robot eye level? Our ultimate goal is to try to be able to square to the bar by detecting the faces on the edge

Here is a shot of what I mean:

[Album] Imgur

Also, does anyone know where to get these LED’s, we want to recreate something like this to test

Does anyone have any pictures from a lower angle of the scale? I am worried from a lower angle close to the nuetral zone (aka a robot near the scale) you won’t be able to see anything

#2

There’s no image

#3

my bad! https://imgur.com/a/XeXGy

Basically just meant what would this platform look like from a lower angle

#4

It’d probably look like a high up row of colored leds…

EDIT: Sorry, if I came off as condescending, that was not my intention at all. I don’t think the lip will affect LEDs much. I think as long as your camera is aimed in the right direction you should be able to get a pretty good view of them. Not sure how to test with that at home though…

#5

was more asking because of that lip, how much of the scale edge will be visible. I haven’t seen it in person, so its difficult to tell how much of it is visible without seeing it in person

Its difficult to tell how much of that visibility translates from the LEDs at a lower angle

you’re probably right, but I can tell you took your chance to be condescending and not think about what I was asking

#6

There may be some folks who attended kickoff who got motion video of the scale lights (which, btw, appear to be identifical to the lights around the scale). If so, hopefully they’ll share.

Short of that:

I’ve asked a rep from FIRST about availability of robot-camera-level video of lights around scales and switches - and the response was the earliest one could reasonably get video of this on a real field (including real lighting scenarios) is week 1 competitions.

Our plan is to put together some opencv sample code for detecting lights that other teams can reference. We wrote similar code (for the Jetson TK1) for detecting angle and distance to the vertical lights in the 2016 Stronghold game (here’s the code, see the process_tower_lights() function). Just one word of caution, this is more complicated (more subject to unwanted impacts from lighting conditions) than vision processing of the retroreflective targets. So my encouragement is to ensure you really need this feature and have the resources to support it before going down the path - it will be most likely hard to get sufficient video sequences to test it until competitions begin. If your team does choose to tackle it, I’ll be happy to collaborate.

If anyone else has any other video sources (photos are ok too) of the lights around the plates on the scales and switches, I’d be grateful if you could share it.

#7

This is awesome! Thank you very much!
If you have them available and could share the camera matrix and the distortion coefficients for the camera which these pictures were taken with that would be even more super helpful. Or at least approximate fx/fy for this 640x360 mode. Please? :smiley:

#8

They are made by Philips, and if you read the game manual::rtm:: you will find the model name in the descriptions of the field components. But be warned, they are quite expensive.

If you want to build a mockup, I would recommend the WS2812B-based strips, driven from an Arduino, or even the non-addressable strips driven from a cheap controller. You can find vendors on eBay who sell these inexpensively. No, your mockup will not be an exact match. It will be plenty good enough!

#9

this is great. Were these shots taken with the lights off?

also, we would love to help with some sort of system to process the scale lighting… we were thinking at the vary least working on squaring to the scale

#10