Procedural generation of simple 3D models of trees

Generating a realistic tree

The length and angle of each branch is random. Initially, the random function was called each time a new branch was created. However, running the profiler showed that approximately 60% of script runtime was due to the random function, as over 300 branches were created for some trees. Therefore, the approach was changed to run the random function once at the start of the script, generating an array of 10000 numbers uniformly distributed between 0 and 1.

MATLAB structures for elegant expansion of model

A challenge was encountered when coding the parameter n_generations. This was due to the iterative nature of branch generation. During the process of branch generation, the co-ordinates of the branches in each successive generation were represented by matrices from p_0 to p_n. The matrix p_1 was produced by a function taking p_0 as an input, and so on up to the iteration specified by n_generations. A solution was necessary to provide the ability to control at which point these iterations would stop. One method considered was as follows:

if n_generations >= 1
p_1 = function(p_0)
if n_generations >= 2
p_2 = function(p_1)
etc…

However, this approach is not elegant and is more importantly not expandable. Furthermore, a solution was required to refer to the matrix representing the final generation of branches, so that leaves could be drawn on the ends of these branches. The solution chosen was to create a structure to which the matrices p_0 to p_n would be successively added. A more concise version of the above code could then be written:

repeat n_generations times
  append array to structure equal to function (last array in structure)

The final generation of branches could be defined as the final array in the structure. Therefore this elegant solution solved two problems simultaneously.

Implementation of pine tree