-
Changing the code structure of the project and also the point clouds processing approach.
A point cloud is represented by a structured numpy array with different fields for each point : X coordinate, Y coordinate, Z coordinate, Blue, Green, Red... The PDAL pipeline written by the user is decomposed in stages and each stage is executed on a numpy array representing a cloud (or a tile in case of single cloud processing).
-
Taking numpy array for each point cloud instead of an entire file.
-
The unmanaged memory generated by a task is now suppressed when the task is finished.
This new approach allows the user to process the number of tasks they want, regardless the number of workers.