Once DDPIV was commercialized and competing technologies such as Tomo-PIV appeared, it became evident that one of DDPIV's advantages was in the possibility for incomparable processing speed.
Reconstruction of the three dimensional state of the fluid volume begins by searching three images for triads of points that form triangles geometrically similar to the layout of the three cameras. Original algorithms were O(n^{3})—the room for improvement was obvious.
I developed a new algorithm that achieved speeds as practically close to the theoretical maximum as possible. Reconstructions were sped up by upwards of 600 times in typical user scenarios (in single-threaded implementations), thus making DDPIV the premier technique for use in in-line processing applications.
The key component of the new algorithm is an easily parallelizable data structure which should make significant improvements through multithreading almost trivial.
The algorithm is currently being implemented into the regular workflow at the California Institute of Technology and INSEAN. |