New video! Relying on RGB data to identify colours doesn’t apply well to real world application, (damn inverse square law!). So This week in The Factory we’ve made code improvements to the PiicoDev VEML6040 Colour Sensor so it’s now super easy to identify objects by their colour. We also find and fix an interesting MOSFET behaviour preventing our high CRI LED from turning off. Then we show off some assembled GlowBit matrix prototypes and discuss potential solutions for easy tiling.
What sort of reading distance are you expecting to get from these?
In the video you’re holding your lunch pretty close to the sensor to get a read…
Would not the properties such as the actual colour of the illuminating LED have an influence on the results from different coloured objects. I think to be meaningful this illumination would have to be close to pure white. Methinks a bit hard to achieve with LEDs.
For reflected light, the target has to be quite close so it doesn’t get swamped by ambient light. In the video I’m testing in a brightly lit office. If the sensor were mounted in an enclosed chamber, perhaps with non-reflective walls, the range would probably be farther - something to try!
Absolutely, yes. The LED we’re using has a high Colour Rendering Index for this reason. High CRI LEDs are used wherever accurate colour is needed like in photography or film lights.
For best results, it’s best to calibrate the sensor for colour classification tasks.
Thanks for that. My fault for not keeping upin my old age. I was not aware these LEDs with more colour accuracy were available although I am not surprised in this day and age.