Fix bug in computing grid size

[Andreas-Piter]

Often the computation did not stop when selecting a grid size smaller than 100m in step 1. This seems to have numerical reasons.

[Andreas-Piter]

@mahmud1, could you please check if it solves the problem that you reported? I tested it locally and seems to work fine.

[mahmud1]

@mahmud1, could you please check if it solves the problem that you reported? I tested it locally and seems to work fine.

@Andreas-Piter, Thank you. I ran the fix on a small dataset with grid_size=20, but it still didn't resolve the issue I had. The memory usage remains very high and the process is killed at some point in step 1.

ascending orbit -> flip up-down
2025-02-27 21:37:12,790 - INFO -     ---------------------------------------------------------------------------------
2025-02-27 21:37:12,790 - INFO -                          STEP 1:     CONSISTENCY CHECK
2025-02-27 21:37:12,791 - INFO -     ---------------------------------------------------------------------------------
2025-02-27 21:37:12,791 - INFO -                               Parameter           value         default
2025-02-27 21:37:12,791 - INFO -                               _________           _____         _______
2025-02-27 21:37:12,791 - INFO -                            coherence_p1             0.9             0.9
2025-02-27 21:37:12,791 - INFO -                               grid_size              20 <---        200
2025-02-27 21:37:12,791 - INFO -                            mask_p1_file            None <---           
2025-02-27 21:37:12,791 - INFO -                  num_nearest_neighbours              30              30
2025-02-27 21:37:12,791 - INFO -                          max_arc_length          999999 <---       None
2025-02-27 21:37:12,791 - INFO -                          velocity_bound             0.3 <---        0.1
2025-02-27 21:37:12,791 - INFO -                         dem_error_bound           500.0 <---      100.0
2025-02-27 21:37:12,791 - INFO -                num_optimization_samples             500 <---        100
2025-02-27 21:37:12,791 - INFO -                arc_unwrapping_coherence             0.6             0.6
2025-02-27 21:37:12,791 - INFO -                             min_num_arc               3               3
2025-02-27 21:37:12,791 - INFO - 
2025-02-27 21:37:12,796 - INFO - reading box None from file: ~/outputs/temporal_coherence.h5 ...
2025-02-27 21:37:13,363 - INFO - 990000 grid cells created.
2025-02-27 21:37:19,844 - INFO - No mask for area of interest given.
2025-02-27 21:37:21,627 - INFO - reading box None from file: ~/outputs/ifg_stack.h5 ...
2025-02-27 21:37:22,039 - INFO - write data to ~/outputs/p1_ifg_wr.h5...
2025-02-27 21:37:22,092 - INFO - create distance matrix between all points...
2025-02-27 21:38:39,485 - INFO - Triangulate points with 30-nearest neighbours.
[==================================================] 60909/60909 points triangulated    3s /     0s
2025-02-27 21:38:42,581 - INFO - Triangulate points with global delaunay.
2025-02-27 21:38:43,461 - INFO - remove arcs with length > 999999.
2025-02-27 21:42:09,012 - INFO - retrieve arcs from adjacency matrix.
2025-02-27 21:42:11,106 - INFO - no. arcs:	1015673
2025-02-27 21:42:25,806 - INFO - ifg arc observations created.
2025-02-27 21:42:25,806 - INFO - write data to ~/outputs/point_network.h5...
2025-02-27 21:42:25,918 - INFO - ########## TEMPORAL UNWRAPPING: AMBIGUITY FUNCTION ##########
2025-02-27 21:42:25,918 - INFO - start parallel processing with 5 cores.
2025-02-27 21:42:39,564 - INFO - Finished temporal unwrapping.
2025-02-27 21:42:39,577 - INFO - write data to ~/outputs/point_network_parameter.h5...
2025-02-27 21:47:22,168 - INFO - Detect points with low quality arcs (mean): < 0.6
2025-02-27 21:47:22,169 - INFO - Removal of points whose arcs are incoherent in average.
zsh: killed     sarvey -f config.json 0 2

[Andreas-Piter]

The computational problem that you are reporting seems not to be an error in the software. However, the number of arcs that need to be checked is huge: >1mio.

2025-02-27 21:42:11,106 - INFO - no. arcs: 1015673

This arises due to

grid_size 20 <--- 200

combined with

num_nearest_neighbours 30 30

You could try again with a smaller number of num_nearest_neighbours, maybe even just 1 (I think 0 is not allowed - but should be implemented in future).