def _dbscan_caller(cluster_candidates):
print("CLUESTER CAND", [p for p in cluster_candidates])
line_seg_index = BestAvailableClusterCandidateIndex(cluster_candidates, epsilon)
print("LINE SSEG INDEX", line_seg_index)
return dbscan(cluster_candidates_index=line_seg_index, #TrajectoryLineSegmentCandidateIndex(cluster_candidates), \
min_neighbors=min_neighbors, \
cluster_factory=TrajectoryClusterFactory())
def _dbscan_caller(cluster_candidates):
line_seg_index = BestAvailableClusterCandidateIndex(cluster_candidates, epsilon)
print("return dbscan")
# print(line_seg_index)
return dbscan(cluster_candidates_index=line_seg_index, #TrajectoryLineSegmentCandidateIndex(cluster_candidates), \
min_neighbors=min_neighbors, \
cluster_factory=TrajectoryClusterFactory())
def run_traclus(trajs, eps, min_lns, min_traj, min_vline, min_prev_dist):
# Cleaning
trajs = [[Point(**pt) for pt in traj] for traj in trajs]
trajs = get_cleaned_trajectories(trajs)
print('Number of trajectories after clean : {}'.format(len(trajs)))
trajs = [Trajectory(traj, tid) for tid, traj in enumerate(trajs)]
# Partitioning
cluster_candidates_tls = []
tls_factory = TrajectoryLineSegmentFactory()
for traj in trajs:
p_indices = call_partition_trajectory(traj.pts)
p_pts = filter_by_indices(p_indices, traj.pts)
traj.p_pts = p_pts
ls_list = get_ls_list(p_pts)
if len(ls_list) <= 0:
raise Exception()
for ls in ls_list:
tls = tls_factory.create(ls, traj.tid)
cluster_candidates_tls.append(tls)
print('Number of segments (cluster candidates) : {}'.format(
len(cluster_candidates_tls)))
# Clustering (DBSCAN)
tls_index = BestAvailableClusterCandidateIndex(cluster_candidates_tls, eps)
tcluster_factory = TrajectoryClusterFactory()
tclusters = dbscan(tls_index, min_lns, tcluster_factory)
print('Number of clusters : {}'.format(len(tclusters)))
# Representative trajectory
rtrajs = []
for tc in tclusters:
if tc.get_num_of_trajs() >= min_traj:
tls_list = tc.get_members()
r_pts = get_rline_pts(tls_list, min_vline, min_prev_dist)
rtraj = RepresentativeTrajectory(r_pts, tc.cid)
rtrajs.append(rtraj)
print('Number of representative trajectories : {}'.format(len(rtrajs)))
result = {
'partitioned_trajectories': trajs,
'clusters': tclusters,
'representative_trajectories': rtrajs
}
return result
def test_dbscan(self):
cluster_factory = ClusterFactory()
for test_ob in self.test_cases:
clusters = dbscan(test_ob['test_points'], test_ob['epsilon'], \
test_ob['min_neighbors'], cluster_factory)
for expected in test_ob['expected_clusters']:
self.assertTrue(self.find_single_matching_cluster(expected, clusters), \
"couldn't find matching cluster in " + str(test_ob) + \
" in clusters: " + str(clusters))
for point in test_ob['test_points']:
self.assertTrue(point.is_classified())
for point in test_ob['expected_noise']:
self.assertTrue(point.is_noise())
for single_cluster in clusters:
for point in single_cluster.members:
self.assertEqual(single_cluster, point.cluster)
self.assertFalse(point.is_noise())
def test_dbscan(self):
cluster_factory = ClusterFactory()
for test_ob in self.test_cases:
clusters = dbscan(test_ob['test_points'], test_ob['epsilon'], \
test_ob['min_neighbors'], cluster_factory)
for expected in test_ob['expected_clusters']:
self.assertTrue(self.find_single_matching_cluster(expected, clusters), \
"couldn't find matching cluster in " + str(test_ob) + \
" in clusters: " + str(clusters))
for point in test_ob['test_points']:
self.assertTrue(point.is_classified())
for point in test_ob['expected_noise']:
self.assertTrue(point.is_noise())
for single_cluster in clusters:
for point in single_cluster.members:
self.assertEqual(single_cluster, point.cluster)
self.assertFalse(point.is_noise())
def _dbscan_caller(cluster_candidates):
return dbscan(cluster_candidates=cluster_candidates, epsilon=epsilon, min_neighbors=min_neighbors, \
cluster_factory=TrajectoryClusterFactory())
def _dbscan_caller(cluster_candidates):
return dbscan(cluster_candidates=cluster_candidates, epsilon=epsilon, min_neighbors=min_neighbors, \
cluster_factory=TrajectoryClusterFactory())
def _dbscan_caller(cluster_candidates):
line_seg_index = BestAvailableClusterCandidateIndex(cluster_candidates, epsilon)
return dbscan(cluster_candidates_index=line_seg_index, #TrajectoryLineSegmentCandidateIndex(cluster_candidates), \
min_neighbors=min_neighbors, \
cluster_factory=TrajectoryClusterFactory())
def dbscan_caller(cluster_candidates):
line_seg_index = TrajectoryLineSegmentCandidateIndex(
cluster_candidates, epsilon)
return dbscan(cluster_candidates_index=line_seg_index, #TrajectoryLineSegmentCandidateIndex(cluster_candidates), \
min_neighbors=min_neighbors, \
cluster_factory=TrajectoryClusterFactory())
