Ignore 0 in cell match (#350)

Since 0 in labels means that no cell was identified in that pixel, we need to
ignore 0 when matching. Otherwise we could have a situation where e.g the
cell overlaps 0.55 with 0 and 0.45 with some cell and it would end up unmatched
(`ID_coverage` would be 0 and a new track would be initialized)

Author: ritvje

pysteps/tracking/tdating.py

@@ -282,6 +282,11 @@ def match(cells_ad, labels):
         ID_vec = labels[cell_a.y, cell_a.x]
         IDs = np.unique(ID_vec)
         n_IDs = len(IDs)
+        if n_IDs == 1 and IDs[0] == 0:
+            cells_ov.t_ID[ID_a] = 0
+            continue
+        IDs = IDs[IDs != 0]
+        n_IDs = len(IDs)
         N = np.zeros(n_IDs)
         for n in range(n_IDs):
             N[n] = len(np.where(ID_vec == IDs[n])[0])