Sign in to follow this  

Question on little part of snow accumulation algorithm by Paul Fearing (SIGGRAPH2000)

This topic is 3596 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

Hi there, currently I am trying to understand the Ph.D. of Paul Fearing: Computer Modelling of Fallen Snow Well, the most of it is clear, but one thing I do not get. It all starts basically with an approximation of you base mesh. Then so called launch sites are added on every triangle to shoot snow particles into the sky. This is done for having control over each face. By flying upwards the snowflakes are checked for intersection with onjects of the scene. After this they reach the sky and contribute to the snow height of it source launch site. No comes the part I do not understand (on the Siggraph Paper its chapter 3.4 (page "41") and in the Ph.D. its chaper 7.6 (page 131)): These are the lines of the Ph.D.:
Quote:
7.6 Locating Particles in the Sky When a launch site reaches the head of the importance queue, it shoots a batch of particles towards the sky. Batch size is user definable, but generally within the order of 10-15 flakes. Particles originate from the launch site's snow surface, potentially reaching the sky plane unimpeded and contributing to the growth of the parent. We use a simple bucketing and filtering scheme to allocate the successful flakes to the total mass of the sky's available snow, while ensuring that small local areas of sky do not over-contribute. This is important, since the number of particles hitting any particular area of the sky may vary dramatically depending on the complexity of the underlying surfaces. We must ensure that a large concentration of flakes (say, directly above a tree), draws the same total snow as would the sky above a sparse at surface. Furthermore, importance ordering implies that not all launch sites shoot the same number of particles. We start by dividing the sky into a grid of constant size buckets, each of area skyarea. When a ake reaches the sky, we locate it within the appropriate bucket. If a flake's represented area is greater than the area skyarea, we place copies of the flake in surrounding buckets, such that each bucket contains an appropriate amount of the ake area - essentially equivalent to spreading a flake's area over the sky, as shown in Figure 7.27. Without such a filtering scheme, a single flake (representing a very large area) might land in adjacent buckets in adjacent turns; because allocated snow is based on the sum of projected flake area, a change in bucket location would greatly affect the snow allocated to nearby akes. Sky buckets only conceptually store individual flakes. For efficiency reasons, each sky bucket actually keeps a list, sorted by launch site, that maintains the total flake area sent from that site. The flake data structure used to fnd and compute intersection is lost (actually, reused for the next launch site) as soon as all flakes in a group have been resolved.
For those who do not want read all this: I understand this in this way: The sky is splittet up in a grad of "sky buckets" od size SIZE_AREA. Every snowflake is responsible for a little bit of surface (this is explained in chapters before) on which a launch site exists to shoot it up to the sky. This "projected area" is then added the sky bucket in which it receives. If its size is bigger than SIZE_AREA the size that is too much is evanly spread of the neighbouring buckets. Then there are some tricks in there to get every snowflake compareable to others. Ok, and then the total mass of the sky is used to determine how much snow every launch site should get. I think all this works, yes, but here my question: why do one need to make it that complicated? Couldn't it be easyer to just store the percentage of how many particles got through to the sky? Use it to multiply it with the desired maximum snow height? Where am I wrong? Thank you very very much. Tom

Share this post


Link to post
Share on other sites
Hi there,
Although I won't be much help with your question, I was wondering if you read his paper describing modelling a snowflake because I can't to wrap my head around how to calculate how many spherical layers are needed for the triangles the compose the snowflake?

Share this post


Link to post
Share on other sites

This topic is 3596 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Sign in to follow this