Oh wow, this is incredible. Circular barcodes are isotropic barcodes. Why haven't I heard of this before? This looks like a proper patentable idea (not some bullshit software patent). A robust physical scanner wouldn't need 3 lasers in different orientations - one would do. And like you say - one scanline is enough to decode it (and also extract 3d orientation at the same time). 🤯 (edit: this is a reply to Tom's thread - I'm still noob at Slack I guess :P)

Thanks. Yeah, the only crappy bit is the grouping of edges, there is a bit of an n^2 algorithmic blow up there that can be improved. Sensors communicate with the CPU via MPI interface one pixel at a time, a few scan lines in parallel, so there is an mechanically sympathetic mapping of the MPI to multi-core parallel processing of scan lines as they leave the sensor without a full frame buffer adding delay. That's the next step! The ardu cam allows you to skip scan lines so you can sample every 10th scan line at 10x the FPS, which is key to making robots be able to servo using optics without being dog slow