Status check. Alright, how far do I have left to go?
All this time I’ve been thinking I’d use reaction-diffusion to generate the cross veins.
Reasons:
- It’ll look cool.
- It’s similar to the idea of signal gradients.
- Reaction-diffusion is well-known for producing a lot of patterns that occur in nature, like the patterns on fish and large, feline predators.
- It would be interesting to see how changing the R-D parameters using the primary veins as a template would affect the patterns generated.
- I can see how it would work.
Downsides:
- It’ll take some exploration to find parameters that work well. Some failed experimentation, no doubt. Uncertain timeline.
- Running the R-D simulations will be time-consuming. Especially if I increase the resolution to something that’ll work well for video.
I’m thinking I might try Poisson Disc Sampling. I can use one of the Poisson Disc Sampling algorithms to find points that are spread out over the wing’s area. Those points will be the centers for calculating the voronoi diagram.
It won’t have the pretty animations I’d get out of doing it with R-D but I think I’ll get to the goal of generating insect wings more quickly.
scipy.stats.qmc.PoissonDisk needs to know the number of points (samples) to calculate and a minimum distance between them. I can probably estimate the number of points from the area of the space between primary veins. Might need some tweaking of parameters there to figure it out, though. Interesting note that with R-D this would be determined automagically.
So I still need to traverse the pixels between the cross-veins in order to calculate the area. Also, I’m going to need the sampling method to reject samples which are too close to the primary veins. That might force me down a custom implementation but that seems doable. We’ll see.
Okay, so:
- Determine the area between each pair of primary veins.
- Use Poisson Disc to find points.
- Draw voronoi pattern.
How will I draw the voronoi pattern?
scipy.spatial.Voronoi can take a list of input points and provide the endpoints for the ridges. That’ll be enough.
For the R-D workflow I was going to export an image from the main_orthoptera.py script I’ve been working on and have a new script which imports it and runs the simulation. But if I’m doing Poisson Disc and Voronoi I think I can just include it all in the same script.
Honestly, I’m a little sad to leave behind R-D. In fact, I’m telling myself I can always come back to it later. But there are loads of other things I’d like to work on and oh so little time!
For flood fill: I no longer need a point within the entire wing, I need points within each pair of primary veins. Probably the easiest way to get these points is to calculate the average of the endpoints of the root segments for a given pair of primary veins. The resulting point should be safely inside.
Now using a flood fill algo from Stack Overflow. Thank you Ted Klein Bergman.