def test_FormationTrack(self):
PD=0.7
PFD=1.e-3
NFC=4
NFA=1000
model_factory = models.SimpleModelFactory(
model = models.KalmanModel,
dT=1.0,
q=0.001
)
sensor = sensors.BaseSensor(
PD=PD,
VC=1.0,
PFA=1e-6,
BNT=0.03
)
# assign 1 track
obs = models.Obs(np.array([1.0, 2.0]), np.eye(2), sensor )
trk = tracks.FormationTrack(obs,model_factory.create(obs))
# assign 2 track
obs = models.Obs(np.array([3.0, 4.0]), np.eye(2), sensor )
trk = tracks.FormationTrack(obs,model_factory.create(obs))
def test_TrackerEvaluator(self):
eval = trackers.TrackerEvaluator(
tracker=trackers.GNN(sensor=sensors.BaseSensor(R=np.diag(
[0.001, 0.001]),
PD=0.7,
VC=1.0,
PFA=1e-6,
BNT=0.03,
y_mins=[-250, -250],
y_maxs=[+250, +250],
y_stps=[1, 1]),
model_factory=models.SimpleModelFactory(
model=models.KalmanModel, dT=1.0, q=1.0),
track_factory=tracks.BaseTrackFactory(
track=tracks.ScoreManagedTrack)),
tgt_list=[
models.SimpleTarget(SD=2,
x0=[0., 0., +0., +1.],
start_time=0.0),
models.SimpleTarget(SD=2,
x0=[0., 10., +1., -1.],
start_time=0.0),
models.SimpleTarget(SD=2,
x0=[10., 10., -1., -1.],
start_time=0.0),
models.SimpleTarget(SD=2,
x0=[10., 0., +0., +1.],
start_time=1.0)
])
# eval.plot_position()
eval.estimate_track_statistics()
def test_SimpleManagedTrack(self):
ND=3
tracker = MockTracker(
sensor=sensors.BaseSensor(
PD=0.7,
VC=1.0,
PFA=1e-6,
BNT=0.03
),
model_factory=models.SimpleModelFactory(
model=models.KalmanModel,
dT=1.0,
q=0.001
)
)
tgt = np.array([0.0, 0.0, 1.0, 1.0])
for k in range(200):
tracker.count = k
if k==0:
obs = models.Obs(tgt[:2], np.eye(2), tracker.sensor )
trk = tracks.SimpleManagedTrack(
obs,
tracker.model_factory.create(obs),
ND=ND
)
if 0 < k <= 100:
tgt[0:2] += tgt[2:]
trk.assign(models.Obs(
# tgt[:2], np.eye(2), tracker.sensor
np.random.multivariate_normal(tgt[:2], np.eye(2)), np.eye(2), tracker.sensor
))
if 100 < k:
trk.unassign(tracker.sensor)
# judge_deletion test
if 0 <= k < 100+ND:
np.testing.assert_equal(trk.judge_deletion(), False)
pass
elif 100+ND <= k:
np.testing.assert_equal(trk.judge_deletion(), True)
pass
# after loop
if False:
# trk.plot_obs_list()
# trk.plot_mdl_list()
plt.show()
def test_SimpleModelFactory(self):
# to_record/from_record check
mf = models.SimpleModelFactory(model=models.KalmanModel, dT=1.0, q=1.0)
mdl1 = mf.create(
models.Obs(y=np.array([1, 0]),
R=np.zeros((2, 2)) + 0.1,
sensor=sensors.BaseSensor()))
series = mdl1.to_record()
mdl2 = mf.create_from_record(series)
def test_GNN(self):
tracker = trackers.GNN(sensor=sensors.BaseSensor(PD=0.7,
VC=1.0,
PFA=1e-6,
BNT=0.03),
model_factory=models.SimpleModelFactory(
model=models.KalmanModel, dT=1.0, q=0.001),
track_factory=tracks.BaseTrackFactory(
track=tracks.ScoreManagedTrack))
tgt_list = [
np.array([-1, 1]),
np.array([20, 36]),
]
tracker.register_scan([models.Obs(y, np.eye(2) * 5) for y in tgt_list])
tgt_list = [
np.array([0, 0]),
np.array([19, 37]),
np.array([40, 50]),
]
tracker.register_scan([models.Obs(y, np.eye(2) * 5) for y in tgt_list])
def test_ScoreManagedTrack(self):
"""Track Score Function
ref) Design and Analysis of Modern Tracking Systems
6.2 Track Score Function
"""
PD=0.7
PFD=1.e-3
NFC=4
NFA=1000
tracker = MockTracker(
sensor=sensors.BaseSensor(
PD=PD,
VC=1.0,
PFA=1e-6,
BNT=0.03
),
model_factory=models.SimpleModelFactory(
model=models.KalmanModel,
dT=1.0,
q=0.001
)
)
tgt = np.array([0.0, 0.0, 1.0, 1.0])
com_list = []
del_list = []
for k in range(200):
tracker.count = k
if k==0:
obs = models.Obs(tgt[:2], np.eye(2), tracker.sensor )
trk = tracks.ScoreManagedTrack(
obs,
tracker.model_factory.create(obs),
PFD=PFD,
alpha=NFC/3600/NFA,
beta=0.1
)
if 0 < k <= 100:
tgt[0:2] += tgt[2:]
obs = models.Obs(
# tgt[:2], np.eye(2), tracker.sensor
np.random.multivariate_normal(tgt[:2], np.eye(2)), np.eye(2), tracker.sensor
)
if np.random.choice([True, False], p=[PD, 1-PD]):
trk.assign(obs)
else:
trk.unassign(tracker.sensor)
if 100 < k:
trk.unassign(tracker.sensor)
com_list.append(trk.judge_confirmation())
del_list.append(trk.judge_deletion())
# after loop
if False:
# trk.plot_obs_list()
# trk.plot_mdl_list()
# trk.plot_scr_list()
plt.plot([i for i in range(200)], np.array(com_list, int) - np.array(del_list, int))
plt.show()
def test_JPDA(self):
"""Calculate Association of Observations by JPDA Method
ref) Design and Analysis of Modern Tracking Systems
6.6.2 Extension to JPDA
"""
tracker = trackers.JPDA(
sensor=sensors.BaseSensor(PD=0.7, VC=1.0, PFA=1e-6, BNT=0.03),
model_factory=models.SimpleModelFactory(
model=models.PDAKalmanModel, dT=1.0, q=0.0),
track_factory=tracks.BaseTrackFactory(track=tracks.PDATrack,
gate=8))
a = 2
x1 = (a**2 + 2 - 2.5) / 2 / a
x2 = (a**2 + 4 - 3) / 2 / a
obs_list = [
models.Obs(y,
np.eye(2) * 0.5) for y in [
np.array([0, 0]),
np.array([a, 0]),
]
]
tracker.register_scan(obs_list)
obs_list = [
models.Obs(y,
np.eye(2) * 0.5) for y in [
np.array([-1, 0]),
np.array([x1, np.sqrt(2 - x1**2)]),
np.array([x2, np.sqrt(4 - x2**2)]),
]
]
expected = np.array([1.0, 9.0, 2.0, 2.5, 4.0, 3.0])
actual = np.array([
dy @ np.linalg.inv(S) @ dy for dy, S in [
trk.model.residual(obs) for obs in obs_list
for trk in tracker.trk_list
]
])
np.testing.assert_almost_equal(actual, expected)
tracker.register_scan(obs_list)
expected = np.array([
6.47e-5,
5.04e-5,
3.06e-5,
1.44e-5,
1.82e-5,
1.11e-5,
8.60e-6,
6.70e-6,
4.10e-6,
2.40e-6,
])
actual = tracker.hyp_price_list
np.testing.assert_almost_equal(np.sort(actual), np.sort(expected))