Enhance the AIS data availability by screening and interpolation

2017·0.51 MB·

by  Daiyong Zhang

2017 4th International Conference on TransportationInformation and Safety (ICTIS), August 8-10, 2017, Banff, Canada
Enhance the AIS data availability by screening and
interpolation
Daiyong Zhang Jia Li
Schoolof Logistic Engineering, Information Center,
Wuhan University of Technology; Changjiang Maritime Safety Administration;
National Engineering Research Center for Water Transport Schoolof Energyand Power Engineering,
Safety,Wuhan University of Technology; Wuhan University of Technology;
Fujian Provincial Key Laboratory of Information Processing Wuhan, Hubei 430063, China
and Intelligent Control (Minjiang University) [email protected]
Wuhan, Hubei 430063, China
[email protected]
Xinglong Liu
National Engineering Research Center for Water Transport
Qing Wu Safety,Wuhan University of Technology;
Schoolof Logistic Engineering, Department of Physics and Electronic Information
Wuhan University of Technology; Engineering,
National Engineering Research Center for Water Transport Minjiang University;
Safety,Wuhan University of Technology; Wuhan, Hubei 430063, China
Wuhan, Hubei 430063, China [email protected]
[email protected]
Wei He
Xiumin Chu School of Economics and Management,
National Engineering Research Center for Water Transport Minjiang University;
Safety,Wuhan University of Technology; Fuzhou,Fujian350108, China
Wuhan, Hubei 430063, China
[email protected]
Abstract—The AIS broadcasts the ship’s navigation data I. INTRODUCTION
automatically and autonomously through VHF band. It plays an
Due to the advantages of large transportation volume, low
important role in collision avoidance and maritime situational
cost, low energy consumption and less pollution, waterway
awareness. However, in inland waterways, it is commonthat the
transportation plays an important role in the comprehensive
AIS data-link was obstructed by the river bank and mountains,
transportation system in China. Researches on
or sometimes the AIS was encountered with the electromagnetic
informationization and intelligent system of shipping
interference. Consequently, the AIS dynamic data is often lost
and mixed with inaccurate, which may lead to misjudgment of management havealways been the concern of the international
the traffic situation. To address this problem, a method was shipping industry, and getting real-time traffic information is
proposed to enhance the availability of AIS data in this paper. the key to build the intelligent shipping system [1].
Firstly, according to the ships’ maneuverability, a set of factors,
Automatic Identification System (AIS) is composed of
such as the moving distance, speed, acceleration, and course
shore station and ship station.The ship station exploitstheGPS
change rate, were designed to screen the inaccurate AIS data.
module to obtain the dynamic information of vessels, such as
Then, the piecewise cubic Hermiteinterpolationandcubicspline
position (presented by longitude and latitude), speed over
interpolationwere employed to restore the AIS data. A real AIS
ground,course over ground, etc. Meanwhile, the pilot input the
trajectory was introduced to validate the accuracy of the two
interpolation methods, which proves that the cubic spline static information, such as the MMSI, call sign, gross tonnage,
interpolationperformanceisbetter than piecewise cubic Hermite draft and dimension, etc. into the ship station with keyboard.
interpolation. Field experiments in Wuhan reach of Yangtze Then, the ship station modulates the dynamic information and
River show that the method proposed in this paper is highly static information into AIS messages and conveys the AIS
effective. The accuracy of the location is 3.5m, the speed accuracy messages to other ship stations around. According to the IMO
is 0.05m/s, and the course accuracy is up to 0.7 degrees, the AIS regulations, the ship station usually sends dynamic AIS
data can be accurately repaired. messages every 6 to 30 seconds when the ship is normal
voyage [2-5]. The shore station receives and decodes the AIS
Keywords—AIS; data availability; ship maneuverability; cubic message transmitted by the ship station, and then displays the
spline interpolation;trajectory restoration information of the ships on VTS screen. Therefore, the AIS
978-1-5386-0437-3/17/$31.00©2017 IEEE
981
2017 4th International Conference on TransportationInformation and Safety (ICTIS), August 8-10, 2017, Banff, Canada
enables the exchange of ship information among vessels and speed, drift track point, unreasonable acceleration and rate of
maritime administrators. It is helpful for maritime regulation turn. It is necessary to screen the raw AIS data to get the
and ship collision avoidance. correct data.In this paper, we designed five rules to screenthe
inaccurate data according to the ship’s maneuverability, as
AIS is an important means to collect ship traffic
shown in follows:
information at present. AIS messages contain large volumes of
information, thus the AIS is an important data source for A. Unreasonalbe stop in moving
analyzing the motion pattern and navigational risk of vessels. The adjacent AIS data is exactly the same sometimes as the
However, the AIS utilize the VHF band communication, which vessel is moving ahead. For this type of inaccurate data, the
makes the data-link of AIS not reliable. Consequently, the raw cleaning rules are as follows. If the speed of the i-th point is
AIS data is not fully available, i.e. the AIS data is not higher than 2 knots, but the coordinate, speed, and course are
completely correct and integrated [6,7]. The incorrect and lost the same as the i-1th point, the i-th data should be deleted. As
AIS data will interfere with maritime regulation, leading to shown in equation (1):
misjudgment of the maritime situation awareness. Meanwhile,
it will decrease the effectiveness of analysis on ship motion v(cid:32) (vi )2(cid:14)(vi )2 (cid:33)2
pattern and traffic flow, which is based on the AIS data. lon lat
Therefore, it is necessary to study the correctness and integrity |lon (cid:16)lon |(cid:32)0
i i(cid:16)1
of AIS data, and identify the inaccurate AIS data and restore |lat (cid:16)lat |(cid:32)0 i(cid:32)2,3,...,m(cid:16)1,m (1)
the lost AISdata. i i(cid:16)1
|vi (cid:16)vi-1 |(cid:32)0
Ma et al.[8,9] conducted studies on the availability of AIS lon lon
information, and proposed approaches for identifying the |vi (cid:16)vi-1|(cid:32)0
inaccurate AIS data based on the DS evidence theory and lat lat
B. Impossible high speed
improved DSmT theory [10,11], however the construction of
evidence in the proposedmethods lack strict reasoning process. Due to the limitation of power engine and navigational
In order to improve the effectiveness of the evidence, Liu et al. rules, the speed of vessels in inland river is less than 16 knot.
[12,13] gave an likelihood based method to construct the Thus, the correct speed range of AIS data is 0~16 knots, as
evidence by statistical analysis on effective prior data samples, shown in equation (2):
and proposed ER rule and the PCR6 rule based methods to
v (cid:32) (vi )2 (cid:14)(vi )2 (cid:33)16(cid:450)i(cid:32)2,3,...,m(cid:16)1,m (2)
identify the inaccurate AIS data. The evidence based methods i lon lat
can give an accurate identification on AIS data, however the
Any AIS data complies with the formula (2) should be
methods seems inefficient, because too many effective data
deleted.
samples should be collected before application. Interpolation
methods are popular in restoring the time series. Liu et al.[14] C. Drift track point
employed the three spline interpolation method to restore the When the track pointis drift away, the distancebetween the
coordinate of lost AIS data. Nguyen [15] utilized piecewise drift point and the adjacent track points will increase.
cubic Hermite interpolation to restore the trajectoryof vessels. Consequently, the average speed will increase as well. If the
However, these researches are competent only if the lost AIS average speed exceeds the maximum speed of the ship, the
data are not consecutive, and the restoration of speed and track point will be considered to be the erroneous data. As
course are not in consideration. Liu et al. [16] introduced the shown in equation (3):
random algorithm into AIS position restoration, but the
randomness of the position is not strong enough to fit the actual (cid:88) (cid:32) (k (x (cid:16)x ))2 (cid:14)(k (y (cid:16)y ))2 (cid:33)16 (3)
i 1 i i(cid:16)1 1 i i(cid:16)1
AIS data.
Where k(i=1,2) is the factor for transferring coordinate to
i
In this paper, we summarized the categories of inaccurate
distance, and k = 96297.6, k = 111194.9 near latitude 30°N.
1 2
AIS data firstly, and proposed a method to screen the
Any track point that complies with equation (3) should be
inaccurate AIS data.Then, a comparative study was conducted
deleted.
to validate the accuracy of improved cubic spline interpolation
method and piecewise cubic Hermite interpolation. D. Unreasonable acceleration
Experiments show that the method proposed in this paper can According to the design specifications of inland ships, the
correctly screen the inaccurate AIS data and restore the lost distance for a ship to accelerate from zero to the design speed
data effectively. is 20 times longer than the lengthof the ship.When the ship is
empty, the distance reduced to 1/2~2/3 times of the original
II. METHOD FORSCREENING DATA
distance. To obtain the maximum acceleration, the minimum
In order to capture the real time AIS data, an AIS shore distance values to 10 times of the lengthof ship.Assuming that
station was set up (base station model is SAAB R40) in the the length of the ship is L, the design speed is V , the time
m
Wuhan section of the Yangtze River. The shore station can from the static accelerate to the design speed is . The
receive the AIS data from vessels 10 kilometers around. Raw travelling distance of the ship can be obtained by formula (4)
AIS data was collected in October 2016, to validate the and shown in Fig.1.The maximum acceleration of the ship can
proposed method. According to the raw AIS data, the be known from formula (5) according to the distance divides
categories of inaccurate AIS data are summarized as follows: velocity transformation formula.
unreasonable stop during moving forward, impossible high
978-1-5386-0437-3/17/$31.00©2017 IEEE
982
2017 4th International Conference on TransportationInformation and Safety (ICTIS), August 8-10, 2017, Banff, Canada
w 360v
V:speed d (cid:32) (cid:32)(cid:32) (8)
Vm wmax tt kkk(cid:83)(cid:83)LL
Wherewis the steering angle of the ship, the unit isdegree;
L is the length of the ship, the unit is meter; t represents the
time, the unit is seconds; v is the ship speed, unit: m/s. For
example, the length of 110 meters for the ship speed of 8knot,
Lm=0.5×tm×Vm the maximum turning rate is 2.1437°/s, where the minimum
value of kis 2.
0 tm T:time (cid:28595)
After getting rid of the wrong AIS data, we can get the
Fig.1 diagram for computing the maximum distance and acceleration correct AIS sequence. However, the original AIS data will be
L (cid:32)10(cid:117)L(cid:32)0.5(cid:117)t (cid:117)V (4) partly missing, which needs to be repaired to get a complete
m m m and correct data. According to the AIS protocol, the time
V V2 interval of class-A ship station does not exceed 10s, the time
a (cid:32) m (cid:32) m (5) interval of class-B ship station does not exceed 30s. In the
max t 20(cid:117)m
m progress of repairing AIS data, we need to interpolate the data
Usually, the ship's maximum speed is 16 knot, so the V is
m according to the time interval, the maximum time interval of
about 8.23 m/s.The length of the ship can be queried from the
10s and 30s.
AIS static data. For a ship with a length of 110m, the
maximum acceleration is 0.03m/s2. It can be seen that the III. METHOD FORDATA RESTORATION
maximum acceleration of any cargo ship can be calculated by After data cleaning, we can obtain the accurate AIS data
(4), (5), any data beyond the maximum acceleration range series. For better interpolation, the time, longitude, latitude,
should be deleted. velocity and course in the AIS data are selected for pre-
E. Unreasonable rate of turn processing, and the raw data is sorted according to the time.
The latitude and longitude of the starting track point are treated
According to the design specification of inland river ships,
as the origin of coordinates. Calculate the time difference,
the maximum diameter of a ship can be obtained by the
speed vector along longitude direction, speed vector along
formula , in which kis the coefficient to measure the
latitude direction between the adjacent AIS data. Setting as
ship maneuverability, usually for the inland river ship, the
time, as longitude, as latitude, as speed, as course
range of k is in 2~4. The maximum swing diameter represents
angle, is speed vector along longitude, and
the maximum variation rate of the ship track.As shown in Fig.
is speed vector along latitude.
2, when the ship is in constant motion, its path is in the
direction of the tangent of the trajectory. Assume that ships
A. Piecewise cubic Hermiteinterpolation
move from D to D , the corresponding time were t and t , the
0 1 0 1
Suppose there are n track points in the track sequence. The
speed v unchanged, the arc angle is w, the maximum rate of
Hermite interpolation interval of longitude for time section
turn can be derived by equation (6), (7) and (8).
( ) is as follows:
v
lon(t )(cid:32)a t3(cid:14)bt2(cid:14)ct (cid:14)d (9)
D0 i i i i i i i i
t0 On the type derivation function on the rate of time,
S
v dlon(t )
v v (cid:32) i (cid:729)3a t2(cid:14)2b t (cid:14)c (10)
w D2 loni d(t ) i i i i i
t1 i
D w In time section(t,t ), the t ,t,lon ,lon ,v ,v are known,
i i+1 1 2 1 2 lon1 lon2
take them in (9) and (10),we can get:
v
lon(t )(cid:32)a t3(cid:14)b t2(cid:14)ct (cid:14)d (cid:32)lon
1 1 1 1 1 11 1 1
lon(t )(cid:32)a t3(cid:14)b t2(cid:14)ct (cid:14)d (cid:32)lon
2 1 2 1 2 1 2 1 2
(11)
v (cid:32)3a t2(cid:14)2b t (cid:14)c
lon1 1 1 1 1 1
v (cid:32)3a t2(cid:14)2b t (cid:14)c
Fig.2. Schematic diagram of the maximum rate of turn for a ship entering a
lon2 1 2 1 2 1
constant cycle
Solving equation (11) to the interval (t , t ) of the Hermite
1 2
D(cid:32)k(cid:117)L,k(cid:143)[2,4] (6) polynomial coefficienta ,b ,c ,d ,so we can get:
1 1 1 1
(cid:167)w(cid:83)D(cid:183) lon(t)(cid:32)a t3(cid:14)b t2(cid:14)ct(cid:14)d
tt (cid:32)tttt (cid:16)ttt (cid:32)(cid:168) (cid:184) v (7) 1 1 1 1
111 00 (cid:169) 360 (cid:185) dlon(t) (12)
v (cid:32) (cid:729)3a t2(cid:14)2b t(cid:14)c
lon1 dt 1 1 1
978-1-5386-0437-3/17/$31.00©2017 IEEE
983
2017 4th International Conference on TransportationInformation and Safety (ICTIS), August 8-10, 2017, Banff, Canada
The Eq. (12) is the interpolation polynomial in interval (cid:170) v (cid:16)v v (cid:16)v (cid:186)
( ).Hence, the longitude and speed along longitude can be (cid:69) (cid:32)3 (1(cid:16)(cid:68)) loni lo(cid:708)n i(cid:16)1(cid:709)(cid:14)(cid:68) lon(i(cid:14)1) loni
(cid:171) (cid:187)
obtained at any time intervalby Eq. (12). Correspondingly, the i (cid:172) 1 h i h (cid:188)
i(cid:16)1 i
latitude and speed along latitude can be obtained as well. For
Then the expression of (16) is:
the original AIS sequence, the two Hermite ( ) using the
(1(cid:16)(cid:68))m (cid:14)2m (cid:14)(cid:68)m (cid:32)(cid:69).
above equations, we can obtain the corresponding interval of n- i i(cid:16)1 i i i(cid:14)1 i
1 interpolation polynomial like(12). We know that .AIS series can be
obtained on the equations (17):
After interpolation,we can get the piecewise functionabout
the time t.Bycombininggthesppeed alongg longitude and latitude,
(cid:173) 2m (cid:14)(cid:68)m (cid:32)(cid:69)(cid:16)(1(cid:16)(cid:68))v(cid:99)
and restored speed is , and the course can 1 1 2 1 1 lon0
(cid:176)
be obtained by Eq.(13): (1(cid:16)(cid:68))m (cid:14)2m (cid:14)(cid:68)m (cid:32)(cid:69)
(cid:176) 1 1 2 2 3 2
(cid:176)
(cid:173) v (cid:174) .... (17)
arcos( lon)(v (cid:33)0)
(cid:176)(cid:176) v lon (cid:176)(1(cid:16)(cid:68) )m (cid:14)2m (cid:14)(cid:68) m (cid:32)(cid:69)
(cid:84)(cid:32)(cid:174) (13) (cid:176) n(cid:16)2 n(cid:16)3 n(cid:16)2 n(cid:16)2 n(cid:16)1 n(cid:16)2
(cid:176)360(cid:16)arcos(vlon)(v (cid:31)0) (cid:176)(cid:175)(1(cid:16)(cid:68)n(cid:16)1)mn(cid:16)2 (cid:14)2mn(cid:16)1 (cid:32)(cid:69)n(cid:16)2 (cid:16)(cid:68)n(cid:16)1vl(cid:99)onn
(cid:176)(cid:175) v lon The coefficient matrix is obtained by solving the equations (18).
B. cubicspline interpolation
After preprocessing the original data, the velocity of the (cid:170) 2 (cid:68) 0 (cid:186)
1
longitude and latitude is obtained. Because the speed in the (cid:171) (cid:187)
1(cid:16)(cid:68) 2 (cid:68)
direction of longitude and latitude, and the longitude and (cid:171) 2 2 (cid:187)
latitude is continuous in time, so we can get a function about A(cid:32)(cid:171) ... ... ... (cid:187) (18)
time for the speed in the direction of longitude, and the (cid:171) (cid:187)
1(cid:16)(cid:68) 2 (cid:68)
velocity in the direction of latitude respectively. Then, (cid:171) n(cid:16)2 n(cid:16)2(cid:187)
the spline function is integrated to obtain the latitude and (cid:171)(cid:172) 0 1(cid:16)(cid:68) 2 (cid:187)(cid:188)
n(cid:16)1
longitude on the speed of the function. So we can obtain
latitude and longitude, and the speed and course at any time. After obtaining the coefficient matrix, we can get a
For the AIS sequence, in each sub segment ,setathree piecewise spline function solution of and on time t.
times , so: Then we can get the corresponding latitude and longitude by
integrating the velocity in the direction of longitude(cid:28435)we can
s (t )(cid:32)v ,s(cid:99)(t )(cid:32)v(cid:99) (cid:32)a (cid:712)s(cid:99)(t )(cid:32)v(cid:99) (cid:32)a
3 i loni 3 0 lon1 lon1 3 i(cid:14)1 lon(i(cid:14)1) lon(i(cid:14)1) obtain the course through the Eq.(13).
Order ,and ,so we can get: IV. EXPERIMENT
t(cid:16)t t(cid:16)t t(cid:16)t t(cid:16)t (14) In order to validate the proposed restore algorithm, we
s3(t)(cid:32)(cid:77)0( h i)vloni(cid:14)(cid:77)1( h i)vlon(i(cid:14)1)+hi(cid:71)0( h i)mi(cid:14)hi(cid:71)1( h i)mi(cid:14)1 choose 163 AIS data of the ship (call sign HAIYOU668,
i i i i MMSI: 413802276, shiplength 120m, Class-Aship station:the
In equation (14)(cid:77)(cid:708)t(cid:709)(cid:32)(t(cid:16)1)2(2t(cid:16)1), so: transmission time interval is 10s) in October 10, 2016 in
0 Wuhan. The coordinates, original speed and acceleration,
(cid:77)(cid:708)t(cid:709)(cid:32)t2((cid:16)2t(cid:14)3)(cid:77)(cid:708)t(cid:709)(cid:32)t2(t(cid:16)1) original course and rate of turn are shown in Fig. 3,Fig. 4, and
1 0
Fig. 5,respectively.
On (14) for the two order, in we have:
(cid:173) v (cid:16)v 4m (cid:14)2m
s(cid:99)(cid:99)(t )(cid:32)6 lon(i(cid:14)1) loni (cid:16) i i(cid:14)1
(cid:176)
(cid:176) 3 i h2 h
(cid:174) i i (15)
(cid:176)s(cid:99)(cid:99)(t )(cid:32)(cid:16)6vlon(i(cid:14)1) (cid:16)vloni (cid:14) 4mi (cid:14)2mi(cid:14)1
(cid:176) 3 i(cid:14)1 h2 h
(cid:175)
i i
In order to ensure the continuity of the two derivative, so:
m (cid:14)2m 2m (cid:14)m
i(cid:16)1 i (cid:14) i i(cid:14)1
h h
i(cid:16)1 i (16)
v (cid:16)v v (cid:16)v
(cid:32)3( loni lo(cid:708)ni(cid:16)1(cid:709)(cid:14) lon(i(cid:14)1) loni)
h2 h2
i(cid:16)1 i
h
Order:(cid:68) (cid:32) i(cid:16)1 , Fig.3.Trajectory of raw AIS data
i h (cid:14)h
i(cid:16)1 i
978-1-5386-0437-3/17/$31.00©2017 IEEE
984
2017 4th International Conference on TransportationInformation and Safety (ICTIS), August 8-10, 2017, Banff, Canada
are fitted to the actual data, which means the inaccurate data
has been deleted.
To verify the validity of the two methods presented in this
paper, we choose 50 correct AIS data, and delete 5 points
manually in a row. Then, the deleted track points are restored
withthe remained45 data, and compared with the original data.
The resultsareshown inFigg.9and Figg.10.
Fig.4. Original speed and acceleration
Fig.9. Repair comparison of Latitude and Longitude
Fig.5. Original courseand rate of turn
There are too many inaccuratie(cid:28678) in the original data, the
length of the ship is 120 meters, the maximum acceleration is
0.028m/s2. In this voyage, the speed is nomore than 10 knots,
the maximum rate of turn is 1.23 (cid:28733)/s. After screening, the
correct dataare shown in Fig.6,Fig.7and Fig.8, respectively.
Fig.10. Repair comparison of Speedand Course
To analyze the accuracy of longitude and latitude, the
inaccurate of longitude and latitude between restored points
and original points are transfer into distance(cid:28609)The average value
and mean square inaccurate of the two kinds of restoration
methods are shown in Table.1 and Table.2.
TABLE I. Mean value of inaccuratein two methods
Distance Speed Course
Hermite 7.301989 0.089212 -0.73175
Cubic spline 6.533656 0.078576 -0.91702
TABLE II. The inaccuratevariance of the two methods
Distance Speed Course
Fig.6AIS trajectory points after cleaning
Hermite 4.004335 0.060271 0.731704
Cubic spline 2.637348 0.042466 0.751654
From the tables, we can conclude that the improved cubic
spline interpolation method performs better. The piecewise
cubic Hermite interpolation restore the latitude and longitude
by construct a unique cubic polynomial. The derivative of the
cubic polynomial about the time is continuous. Hence the
speed is relatively smooth. But the change of speed wasn’t
Fig.7.Speed and acceleration after cleaning
taken into consideration, so the speed is prone to change
suddenly in each interval. Therefore, the fitting effect is poor.
On contrary, the derivative of the improved cubic spline
interpolation about the time is smooth. The position inaccurate
isabout6.5m, the speed inaccurateis about 0.07m/s, the angle
inaccurateisabout0.91°.
After cleaning the original AIS data with inaccurate, the
AIS sequences were restored by the improved cubic spline
Fig.8.Courseand rate of turnaftercleaning interpolation.The lost data after cleaning is from 100thto 115th.
After data cleaning, 110 correct data are obtained. The The experimental results are shown in Fig.11, Fig.12 and
speed, acceleration, course and rate of turn of the cleaned data Fig.13, respectively.
978-1-5386-0437-3/17/$31.00©2017 IEEE
985
2017 4th International Conference on TransportationInformation and Safety (ICTIS), August 8-10, 2017, Banff, Canada
(Grant No. 2015J05108) and Open Fund Project of Fujian
Provincial Key Laboratory of Information Processing and
Intelligent Control (Minjiang University) (No. MJUKF201727).
The authors would like to express their sincere gratitude to
anonymous referees for their valuable comments.
REFERENCES
[1] YAN Xin-ping, MA Feng, CHU Xiu-min, et al. Key technology of
collecting traffic flow on the Yangtze River in Real-Time[J]. Navigation
Fig.11.Trackpointsbefore and after restoration
of China, 2010, 33:40-45.
[2] Pallotta G, Vespe M, Bryan K. Vessel Pattern Knowledge Discovery
from AIS Data: A Framework for Anomaly Detection and Route
Prediction[J]. Entropy, 2013,15(6):2218-2245.
[3] Steven Mascaro K B K A. Learning Abnormal Vessel Behaviour from
AIS Data with Bayesian Networks at Two Time Scales[J]. Tracks a
Journal of Artists Writings, 2010.
[4] Talavera A, Aguasca R, Galván B, et al. Application of Dempster–
Shafer theory for the quantification and propagation of the uncertainty
caused by the use of AIS data[J]. Reliability Engineering & System
Fig.12. The speed and acceleration after restoration
Safety, 2013,111:95-105.
[5] Aarsæther K G, Moan T. Computer vision and ship traffic analysis:
Inferring maneuver patterns from the automatic identification system[M].
CRC Press, 2012.
[6] Michael Baldauf K B F M. Aspects of Technical Reliability of
Navigation Systems and Human Element in Case of Collision
Avoidance: Proceedings of the Navigation Conference and Exhibition,
London, UK., 2014[C].
[7] Zhu Feixiang, Zhang Yingjun, Gao Jiang. Research on [J]. ship behavior
data mining based on China navigation, 2012 (02): 50-54.
Fig.13.The courseand rateof turn after restoration
[8] Ma Feng, Yan Xinping, min Xiu min, et al. Correlation between signal
V. CONCLUSION failure and field strength of ship automatic identification system [J].
Journal of Dalian Maritime University, 2011 (03): 111-114.
In this paper, five categories of inaccurate in AIS data are
[9] CHU Xiu-min, LIU Tong, MA Feng, et.al. Distribution characteristic of
presented, and correspondingly, we designed five rules to clean
AIS signal field intensity along mountainous waterway[J]. Journal of
the inaccurate according to the ships’ maneuverability. Field Traffic and Transportation Engineering,2014,14(06):117-126.
experiment indicates that the five rules could delete the
[10] MA Feng, CHU Xiu-min, YAN Xin-ping, et al. InaccurateDistinguish
inaccurate data effectively. In order to restore the deleted and of AIS Based on Evidence Combination[J]. Journal of theoretical and
lost AIS data, piecewise cubic Hermite interpolation and cubic applied information technology, 2012, 45(1):83-90.
spline interpolation are employed. And we use real AIS [11] MA Feng, CHU Xiu-min, LIU Chen-guang.The inaccurate
trajectories to validate the accuracy of the two interpolation distinguishing of automatic identification system based on improved
evidence similarity[C]IEEE. 2th International Conference on
methods. The result shows that the cubic spline interpolation
Transportation Information and Safety. Wuhan, China. IEEE, 2013:715-
can restore the fragmentaryAIS data better.
722
However, the cubic spline interpolation is effective only [12] Liu Xinglong, Yi Xiu min, Ma Feng, et al. Abnormal dynamic
information screening method for AIS message [J]. Journal of traffic and
when the number of continuouslylost data is less than five. In
transportation engineering, 2016 (05): 142-150.
fact, more data will continuously lost when the ship passing
[13] Liu Xinglong, Yi Xiu min, Ma Feng, et al. Research on the method of
through the curve channel and mountainous waterway, or
PCR6 target fusion recognition based on AIS [J]. Journal of Wuhan
encountering with the electromagnetic interference. In this University of Technology, 2015 (08): 42-50.
paper, wedidn’tconsider the situationthat more than five track [14] Liu Liqun, Wu Chaozhong, Zhu Duanfeng, et al. Study on ship
points are lost. In the future work, we will take this situation trajectory restoration based on Vondrak filtering and three spline
into consideration. interpolation [J]. traffic information and safety, 2015 (4): 100-105.(cid:676)in
Chinese(cid:677)
ACKNOWLEDGMENT [15] Nguyen V, Im N, Lee S. The Interpolation Method for the missing AIS
The present study is financial supported by the Double Data of Ship[J]. Journal of Navigation and Port Research,
2015,39(5):377-384.
First-rate Project of WUT, the National Natural Science
[16] Liu C, Chen X. Inference of Single Vessel Behaviour with Incomplete
Foundation of China (No.51479155, No.61273234), the
Satellite-based AIS Data[J]. 2013,66(6):813.
Excellent Dissertation Cultivation Funds of Wuhan University
of Technology,Natural Science Foundation of Fujian Province
978-1-5386-0437-3/17/$31.00©2017 IEEE
986