Trajectory Clustering in an Intersection by GDTW

Algorithm 5

(ADPC_allocate_sample).

	ADPC_allocate_sample
	Input: D: normalized asymmetric generalized DTW
	centeridxs: candidate cluster center index list
	thre_c2m: threshold from center to member
	thre_m2c: threshold from member to center
	Return: centeridx_to_memberidxs. outlieridxs:
	outlier index list.
(1)	cm_mat = D[centeridxs,:]
(2)	mc_mat = D[:, centeridxs]
(3)	outlieridxs = []
(4)	centeridx_to_memberidxs = dict((i, []) for i in centeridxs)
(5)	for i = 0…len(D)-1:
(6)	if i in centeridx_to_memberidxs:
(7)	centeridx_to_memberidxs[i].append(i)
(8)	continue
(9)	br_c2m = cm_mat[:, i] < thre_c2m
(10)	br_m2c = mc_mat[i] < thre_m2c
(11)	ii = true_idxs(br_c2m & br_m2c)
(12)	if len(ii) = = 0:
(13)	outlieridxs.append(i)
(14)	continue
(15)	cj = argmin(cm_mat[j, i] + mc_mat[i, j] for j in ii)
(16)	ci = centeridxs[ ii[cj] ]
(17)	centeridx_to_memberidxs[ci].append(i)
(18)	return centeridx_to_memberidxs, outlieridxs