CN109001722A - A kind of ship track data fusion method based on LSTM model - Google Patents
A kind of ship track data fusion method based on LSTM model Download PDFInfo
- Publication number
- CN109001722A CN109001722A CN201810537118.6A CN201810537118A CN109001722A CN 109001722 A CN109001722 A CN 109001722A CN 201810537118 A CN201810537118 A CN 201810537118A CN 109001722 A CN109001722 A CN 109001722A
- Authority
- CN
- China
- Prior art keywords
- time
- acquisition
- coordinate information
- information
- ship
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/02—Systems using reflection of radio waves, e.g. primary radar systems; Analogous systems
- G01S13/50—Systems of measurement based on relative movement of target
- G01S13/58—Velocity or trajectory determination systems; Sense-of-movement determination systems
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/86—Combinations of radar systems with non-radar systems, e.g. sonar, direction finder
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S19/00—Satellite radio beacon positioning systems; Determining position, velocity or attitude using signals transmitted by such systems
- G01S19/38—Determining a navigation solution using signals transmitted by a satellite radio beacon positioning system
- G01S19/39—Determining a navigation solution using signals transmitted by a satellite radio beacon positioning system the satellite radio beacon positioning system transmitting time-stamped messages, e.g. GPS [Global Positioning System], GLONASS [Global Orbiting Navigation Satellite System] or GALILEO
- G01S19/42—Determining position
- G01S19/45—Determining position by combining measurements of signals from the satellite radio beacon positioning system with a supplementary measurement
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Traffic Control Systems (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention proposes a kind of ship track data fusion methods based on LSTM model.The present invention collects the ship's navigation track of ship AIS system, shipborne radar and high-precision GPS acquisition in primary navigation as historical data, pass through dynamic latitude and longitude coordinates information of AIS system acquisition ship during one section of navigation, dynamic image coordinate information of ship during one section of navigation is acquired by shipborne radar, as to training data；Radar latitude and longitude coordinates information, AIS coordinate information and high-precision GPS latitude and longitude coordinates information are normalized respectively；It is grouped shipborne radar, AIS system and the ship's navigation trajectory coordinates information of high-precision GPS acquisition after normalization to obtain grouping coordinate information；Radar and AIS coordinate information after being grouped by normalization carry out initial training to LSTM model as training set, are then adjusted using by being grouped normalized high-precision GPS coordinate information to the weight parameter of LSTM model.The invention has the advantages that improving ship track fusion accuracy.
Description
Technical field
The present invention relates to ship information perception and the field of merging, and particularly relate to a kind of ship based on LSTM model
Track data fusion method.
Background technique
The whole world has entered " industry 4.0 " developing period, and has done emphasis to high performance ship, ship intelligence manufacture etc. and said
It is bright.Nowadays rapid economic development, marine ships are rapidly developed towards technicalization, rapid direction, and ships quantity is especially transported
The quantity of defeated dangerous material ship is increasing, and ship density is continuously increased, and the safety that above-mentioned reason directly results in maritime traffic is hidden
Trouble is sharply increased, and marine accident occurs frequently, this not only makes lives and properties by significant threat, but also largely causes
The environmental pollution of ocean.In order to ensure navigation safety at sea, shipping management efficiency is improved, the requirement to communication navigation set is not
It is disconnected to improve.
Modern radar technology ARPA (Automatic Radar Plotting Aids, automatic radar plotting aid) or TT
(Target Tracking, target following) can enroll and track automatically target, can provide the multidate information such as ship of target
The information such as position, the speed of a ship or plane, course can provide the display of more comprehensive traffic situation image, to obtain distance between ship.?
In practical application, radar carries out target acquisition by the echo signal information of analysis ship, and there are certain detection blind areas.Separately
Outside, due to working in microwave band, it is easy the interference by terrain environment and meteorological condition, and precision is limited, in fact it could happen that mesh
Target missing inspection and false target phenomenon.The information that AIS is provided is relatively abundant, it can provide real-time dynamic information for example accommodation, the speed of a ship or plane,
The information such as course, UTC (world concordant time).AIS uses radio propagation, works in very high frequency(VHF) VHF (Very High
Frequency) marine frequency range, since radio wave diffraction ability is strong, its operating distance is far, secondly as location data comes from
In Global Satellite Navigation System GNSS (Global NavigationSatellite System), thus its work not by
Distance, position influence, it is often more important that it is hardly limited by weather, provide contain much information and also precision is quite high: and
Since the MMSI of each ship distribution is different, so even if in ship intensively and when target range is closer, will not occur mistake with
The phenomenon that track or leakage tracking.But there is also disadvantages by AIS, since the characteristics of equipment is passively to receive information, it is not
It can actively go to obtain specific objective information；The vessel position that AIS is provided is coordinate points, it can not obtain the video letter of target
Breath and ambient condition information.
It is thus impossible to which AIS is using as the collision avoidance aids that uniquely navigates in navigation.We use AIS equipment backup radar, and will
The fused information of the two is shown on sea chart, this can not only avoid the occurrence of extremely complex display picture, realizes radar mesh
Target automatic identification, the identification for solving operator is difficult, and the data after the fusion of the two information are also more accurate, reliably
Property is higher.The mutual supplement with each other's advantages of this radar and AIS information, so that maritime bridge has obtained a greater degree of guarantee.
Summary of the invention
In order to solve the above-mentioned technical problem, the invention proposes a kind of ship track data fusion sides based on LSTM model
Method.
The technical scheme is that a kind of ship track data fusion method based on LSTM model, this method include with
Lower step:
Step 1: collecting the ship ship's navigation rail that AIS system, shipborne radar and high-precision GPS acquire in primary navigation
Mark passes through ship by dynamic latitude and longitude coordinates information of AIS system acquisition ship during one section of navigation as historical data
Dynamic image coordinate information of radar acquisition ship during one section of navigation is carried, as to training data；
Step 2: by radar latitude and longitude coordinates information, AIS coordinate information and the high-precision GPS latitude and longitude coordinates in step 1
Information is normalized respectively；
Step 3: by the shipborne radar after being normalized in step 2, the ship's navigation rail of AIS system and high-precision GPS acquisition
Every five time points obtain being grouped coordinate information for one group mark coordinate information sequentially in time；
Step 4: radar and AIS coordinate information after being grouped by normalization carry out LSTM model as training set preliminary
Then training uses and passes through Feedback error pair by being grouped normalized high-precision GPS coordinate information as test set
The weight parameter of LSTM model is adjusted trained after LSTM model.
Preferably, acquiring the underway dynamic latitude and longitude coordinates information of ship by AIS described in step 1, need simultaneously
The image information that radar acquires is converted into the coordinate information to match with AIS information, and is arranged sequentially in time respectively
Column track coordinate information；
The dynamic location information of the ship of AIS system acquisition described in step 1 is dynamic latitude and longitude coordinates information:
Stime+i*Δt=(longtime+i*Δt,lattime+i*Δt)i∈[0,N]
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points, Stime+i*ΔtBetween the i-th acquisition
Every dynamic latitude and longitude coordinates information, longtime+i*ΔtFor the longitude coordinate information of the i-th acquisition interval, lattime+i*ΔtIt is i-th
The latitude coordinate information of acquisition interval；
The dynamic location information that shipborne radar described in step 1 acquires ship is that image information is converted by Mercator projection
As the latitude and longitude coordinates information to match with AIS dynamic location information:
Ltime+i*Δt=(xtime+i*Δt,ytime+i*Δt)i∈[0,N]
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points, Ltime+i*ΔtBetween the i-th acquisition
Every dynamic image coordinate information, xtime+i*ΔtFor the i-th acquisition interval longitude coordinate information, ytime+i*ΔtFor the i-th acquisition interval
Latitude coordinate information；
High-precision GPS described in step 1 acquires the dynamic location information of ship as high-precision latitude and longitude coordinates information:
Gtime+i*Δt=(ptime+i*Δt,qtime+i*Δt)i∈[0,N]
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points, Gtime+i*ΔtBetween the i-th acquisition
Every Dynamic High-accuracy image coordinate information, ptime+i*ΔtFor the longitude coordinate information of the i-th acquisition interval, qtime+i*ΔtIt is adopted for i-th
Collect the latitude coordinate information at interval；
Preferably, grouping radar latitude and longitude coordinates information described in step 2 is normalized are as follows:
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points,After normalization
The radar fix information of i-th acquisition interval；
AIS coordinate information described in step 2 is normalized are as follows:
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points,After normalization
The AIS coordinate information of i-th acquisition interval；
High-precision GPS coordinate information after being grouped described in step 2 is normalized are as follows:
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points,After normalization
The coordinate information of the high-precision GPS acquisition of i-th acquisition interval；
Preferably, coordinate information every five sequentially in time that the radar described in step 3 after normalizing acquires
Time point is one group are as follows:
Wherein, tim is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points,For
N-th of coordinate information, WL in jth groupjCoordinate information is grouped for the jth after normalization；
Coordinate information described in step 3 Jing Guo normalized AIS system acquisition every five time points sequentially in time
It is one group are as follows:
Wherein, tim is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points,For
N-th of coordinate information, WS in jth groupjCoordinate information is grouped for the jth after normalization；
By normalized high-precision GPS coordinate information, every five time points are one sequentially in time described in step 3
Group are as follows:
Wherein, tim is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points,
For n-th of coordinate information in jth group, WGjCoordinate information is grouped for the jth after normalization；
Preferably, LSTM model described in step 4, which is one, has α input layer, β hidden layer, γ output
The structure of layer, wherein LSTM learning process focuses on hidden layer using the distinctive cell unit of LSTM, which has input
Door forgets door and out gate；
Input gate indicates whether that new data information is allowed to be added in currently hiding node layer, allows if value is 1 defeated
Enter, does not allow if value is 0；
Forget door and indicate whether that the historical data for retaining current hiding node layer storage retains if value is that 1 i.e. door is opened,
It is closed if it is 0 i.e. door, empties the historical data that present node is stored；
Out gate indicates whether to export present node output valve to next layer, i.e., next hidden layer or output layer,
It is opened if it is 1 i.e. door, the output valve of present node will act on next layer, close if it is 0 i.e. door, present node output valve
It does not export；
LSTM model described in step 4 is that crucial formula is as follows:
xtime+(i+1)*Δt=[WLj,WSj
htime+i*Δt=[WLj,WSj,WGj
ftime+(i+1)*Δt=σ (Wf·[Ctime+i*Δt,htime+i*Δt,xtime+(i+1)*Δt]+bf)
itime+(i+1)*Δt=σ (Wi·[Ctime+i*Δt,htime+i*Δt,xtime+(i+1)*Δt]+bi)
Otime+(i+1)*Δt=σ (Wo·[Ctime+(i+1)*Δt,htime+i*Δt,xtime+(i+1)*Δt]+bo)
Ctime+(i+1)*Δt=ftime+(i+1)*Δt*Ctime+i*Δt+itime+(i+1)*Δt
*tanh(Wc[htime+i*Δt,xtime+(i+1)*Δt]+bc)
htime+(i+1)*Δt=Otime+(i+1)*Δttanh(Ctime+(i+1)*Δt)
Wherein, its concrete meaning is as follows in above-mentioned formula, and f indicates that forgeing door, i expression input gate, O indicates out gate, C table
Show that unit activating vector, h indicate to hide layer unit, WfTo forget the weight matrix between door and other input vectors, WiFor input
Door and other receive the weight matrix between the vector of input data, WoFor the weight square between data before out gate and out gate
Battle array, WcFor the weight matrix between other vectors of unit activating vector, activation primitive is tanh, bf、bi、bo、bcRespectively forget door,
The deviation of input gate, out gate, unit activating vector, different sampling moment indicate with subscript t, wherein htime+(i+1)*Δt's
Value is the hidden layer h by the last momenttime+i*ΔtWith new input xtime+(i+1)*ΔtAnd the oneself state of cell last moment
Ctime+i*Δt, controlled by door compound come in the h at the last one momenttime+(i+1)*ΔtAs export.
Compared with prior art, the present invention reaches information content loss reduction, and processing information content is maximum, and fusion performance is best
Effect, to preferably determine the real trace of ship movement, the method for the present invention combination multisensor carries out data fusion, thus
Spatiotemporal coverage area is extended, detection performance is improved, improves resolving power, improves the reliability of system.
Detailed description of the invention
Fig. 1: principle of the invention flow chart；
Fig. 2: cell entity relationship diagram in the LSTM of addition peephole structure；
Cell cellular construction schematic diagram in Fig. 3: LSTM.
Specific embodiment
Understand for the ease of those of ordinary skill in the art and implement the present invention, with reference to the accompanying drawings and embodiments to this hair
It is bright to be described in further detail, it should be understood that implementation example described herein is merely to illustrate and explain the present invention, not
For limiting the present invention.
Introduce embodiments of the present invention below with reference to Fig. 1 to Fig. 3, embodiment of the present invention specifically includes the following steps:
Step 1: collecting the ship ship's navigation rail that AIS system, shipborne radar and high-precision GPS acquire in primary navigation
Mark passes through ship by dynamic latitude and longitude coordinates information of AIS system acquisition ship during one section of navigation as historical data
Dynamic image coordinate information of radar acquisition ship during one section of navigation is carried, as to training data；
The underway dynamic latitude and longitude coordinates information of ship is acquired by AIS described in step 1, while needing to adopt radar
The image information of collection is converted into the coordinate information to match with AIS information, and arranges track coordinate sequentially in time respectively
Information；
The dynamic location information of the ship of AIS system acquisition described in step 1 is dynamic latitude and longitude coordinates information:
Stime+i*Δt=(longtime+i*Δt,lattime+i*Δt)i∈[0,N]
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points, Stime+i*ΔtBetween the i-th acquisition
Every dynamic latitude and longitude coordinates information, longtime+i*ΔtFor the longitude coordinate information of the i-th acquisition interval, lattime+i*ΔtIt is i-th
The latitude coordinate information of acquisition interval；
The dynamic location information that shipborne radar described in step 1 acquires ship is that image information is converted by Mercator projection
As the latitude and longitude coordinates information to match with AIS dynamic location information:
Ltime+i*Δt=(xtime+i*Δt,ytime+i*Δt)i∈[0,N]
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points, Ltime+i*ΔtBetween the i-th acquisition
Every dynamic image coordinate information, xtime+i*ΔtFor the i-th acquisition interval longitude coordinate information, ytime+i*ΔtFor the i-th acquisition interval
Latitude coordinate information；
High-precision GPS described in step 1 acquires the dynamic location information of ship as high-precision latitude and longitude coordinates information:
Gtime+i*Δt=(ptime+i*Δt,qtime+i*Δt)i∈[0,N]
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points, Gtime+i*ΔtBetween the i-th acquisition
Every Dynamic High-accuracy image coordinate information, ptime+i*ΔtFor the longitude coordinate information of the i-th acquisition interval, qtime+i*ΔtIt is adopted for i-th
Collect the latitude coordinate information at interval；
Step 2: by radar latitude and longitude coordinates information, AIS coordinate information and the high-precision GPS latitude and longitude coordinates in step 1
Information is normalized respectively；
Grouping radar latitude and longitude coordinates information described in step 2 is normalized are as follows:
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points,After normalization
The radar fix information of i-th acquisition interval；
AIS coordinate information described in step 2 is normalized are as follows:
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points,After normalization
The AIS coordinate information of i-th acquisition interval；
High-precision GPS coordinate information after being grouped described in step 2 is normalized are as follows:
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points,After normalization
The coordinate information of the high-precision GPS acquisition of i-th acquisition interval；
Step 3: by the shipborne radar after being normalized in step 2, the ship's navigation rail of AIS system and high-precision GPS acquisition
Every five time points obtain being grouped coordinate information for one group mark coordinate information sequentially in time；
Every five time points are one to the coordinate information of radar acquisition described in step 3 after normalizing sequentially in time
Group are as follows:
Wherein, tim is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points,For
N-th of coordinate information, WL in jth groupjCoordinate information is grouped for the jth after normalization；
Coordinate information described in step 3 Jing Guo normalized AIS system acquisition every five time points sequentially in time
It is one group are as follows:
Wherein, tim is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points,For
N-th of coordinate information, WS in jth groupjCoordinate information is grouped for the jth after normalization；
By normalized high-precision GPS coordinate information, every five time points are one sequentially in time described in step 3
Group are as follows:
Wherein, tim is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points,
For n-th of coordinate information in jth group, WGjCoordinate information is grouped for the jth after normalization；
Step 4: passing through the radar and AIS coordinate information WL after normalization groupingj, WSjAs training set to LSTM model into
Row initial training, then using by being grouped normalized high-precision GPS coordinate information WGjIt is reversed by error as test set
Transmitting the weight parameter of LSTM model is adjusted trained after LSTM model；
LSTM model described in step 4, which is one, has α=12 input layer, and β=18 hidden layer, γ=4 are defeated
The structure of layer out, wherein LSTM learning process focuses on hidden layer using the distinctive cell unit of LSTM, which has defeated
Introduction forgets door and out gate；
Input gate indicates whether that new data information is allowed to be added in currently hiding node layer, allows if value is 1 defeated
Enter, does not allow if value is 0；
Forget door and indicate whether that the historical data for retaining current hiding node layer storage retains if value is that 1 i.e. door is opened,
It is closed if it is 0 i.e. door, empties the historical data that present node is stored；
Out gate indicates whether to export present node output valve to next layer, i.e., next hidden layer or output layer,
It is opened if it is 1 i.e. door, the output valve of present node will act on next layer, close if it is 0 i.e. door, present node output valve
It does not export；
LSTM model described in step 4 is that crucial formula is as follows:
xtime+(i+1)*Δt=[WLj,WSj]
htime+i*Δt=[WLj,WSj,WGj]
ftime+(i+1)*Δt=σ (Wf·[Ctime+i*Δt,htime+i*Δt,xtime+(i+1)*Δt]+bf)
itime+(i+1)*Δt=σ (Wi·[Ctime+i*Δt,htime+i*Δt,xtime+(i+1)*Δt]+bi)
Otime+(i+1)*Δt=σ (Wo·[Ctime+(i+1)*Δt,htime+i*Δt,xtime+(i+1)*Δt]+bo)
Ctime+(i+1)*Δt=ftime+(i+1)*Δt*Ctime+i*Δt+itime+(i+1)*Δt
*tanh(Wc[htime+i*Δt,xtime+(i+1)*Δt]+bc)
htime+(i+1)*Δt=Otime+(i+1)*Δttanh(Ctime+(i+1)*Δt)
Wherein, its concrete meaning is as follows in above-mentioned formula, and f indicates that forgeing door, i expression input gate, O indicates out gate, C table
Show that unit activating vector, h indicate to hide layer unit, WfTo forget the weight matrix between door and other input vectors, WiFor input
Door and other receive the weight matrix between the vector of input data, WoFor the weight square between data before out gate and out gate
Battle array, WcFor the weight matrix between other vectors of unit activating vector, activation primitive is tanh, bf、bi、bo、bcRespectively forget door,
The deviation of input gate, out gate, unit activating vector, different sampling moment indicate with subscript t, wherein htime+(i+1)*Δt's
Value is the hidden layer h by the last momenttime+i*ΔtWith new input xtime+(i+1)*ΔtAnd the oneself state of cell last moment
Ctime+i*Δt, controlled by door compound come in the h at the last one momenttime+(i+1)*ΔtAs export.
The ship's navigation track data input of AIS system to be fused, shipborne radar and high-precision GPS acquisition has been trained
At LSTM model in, LSTM model treats the study of fused data after training, finally obtain fusion complete practical ship
Ship motion.
It should be understood that the part that this specification does not elaborate belongs to the prior art.
It should be understood that the above-mentioned description for preferred embodiment is more detailed, can not therefore be considered to this
The limitation of invention patent protection range, those skilled in the art under the inspiration of the present invention, are not departing from power of the present invention
Benefit requires to make replacement or deformation under protected ambit, fall within the scope of protection of the present invention, this hair
It is bright range is claimed to be determined by the appended claims.
Claims (5)
1. a kind of ship track data fusion method based on LSTM model, which comprises the following steps:
Step 1: collecting ship ship's navigation track of AIS system, shipborne radar and high-precision GPS acquisition in primary navigation and make
Boat-carrying thunder is passed through by dynamic latitude and longitude coordinates information of AIS system acquisition ship during one section of navigation for historical data
Up to dynamic image coordinate information of acquisition ship during one section of navigation, as to training data；
Step 2: by radar latitude and longitude coordinates information, AIS coordinate information and the high-precision GPS latitude and longitude coordinates information in step 1
It is normalized respectively；
Step 3: the ship's navigation track of the shipborne radar after normalizing in step 2, AIS system and high-precision GPS acquisition is sat
Every five time points obtain being grouped coordinate information for one group mark information sequentially in time；
Step 4: radar and AIS coordinate information after being grouped by normalization tentatively instruct LSTM model as training set
Practice, then uses and pass through Feedback error to LSTM by being grouped normalized high-precision GPS coordinate information as test set
The weight parameter of model is adjusted trained after LSTM model.
2. the ship track data fusion method according to claim 1 based on LSTM model, which is characterized in that step 1
Described in by AIS acquire the underway dynamic latitude and longitude coordinates information of ship, while need by radar acquire image information
It is converted into the coordinate information to match with AIS information, and arranges track coordinate information sequentially in time respectively；
The dynamic location information of the ship of AIS system acquisition described in step 1 is dynamic latitude and longitude coordinates information:
Stime+i*Δt=(longtime+i*Δt,lattime+i*Δt)i∈[0,N]
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points, Stime+i*ΔtFor the i-th acquisition interval
Dynamic latitude and longitude coordinates information, longtime+i*ΔtFor the longitude coordinate information of the i-th acquisition interval, lattime+i*ΔtFor the i-th acquisition
The latitude coordinate information at interval；
The dynamic location information of the acquisition ship of shipborne radar described in step 1 is that image information is converted by Mercator projection
The latitude and longitude coordinates information to match with AIS dynamic location information:
Ltime+i*Δt=(xtime+i*Δt,ytime+i*Δt)i∈[0,N]
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points, Ltime+i*ΔtFor the i-th acquisition interval
Dynamic image coordinate information, xtime+i*ΔtFor the i-th acquisition interval longitude coordinate information, ytime+i*ΔtFor the latitude of the i-th acquisition interval
Coordinate information；
High-precision GPS described in step 1 acquires the dynamic location information of ship as high-precision latitude and longitude coordinates information:
Gtime+i*Δt=(ptime+i*Δt,qtime+i*Δt)i∈[0,N]
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points, Gtime+i*ΔtFor the i-th acquisition interval
Dynamic High-accuracy image coordinate information, ptime+i*ΔtFor the longitude coordinate information of the i-th acquisition interval, qtime+i*ΔtBetween the i-th acquisition
Every latitude coordinate information.
3. the ship track data fusion method according to claim 1 based on LSTM model, which is characterized in that step 2
Described in grouping radar latitude and longitude coordinates information be normalized are as follows:
i∈[0,N]
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points,It is adopted for i-th after normalization
Collect the radar fix information at interval；
AIS coordinate information described in step 2 is normalized are as follows:
i∈[0,N]
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points,For i-th after normalization
The AIS coordinate information of acquisition interval；
High-precision GPS coordinate information after being grouped described in step 2 is normalized are as follows:
i∈[0,N]
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points,It is adopted for i-th after normalization
The coordinate information of the high-precision GPS acquisition at collection interval.
4. the ship track data fusion method according to claim 1 based on LSTM model, which is characterized in that step 3
Described in after normalizing radar acquisition coordinate information sequentially in time every five time points be one group are as follows:
j∈[0,N-4]
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points,For jth
N-th of coordinate information in group, WLjCoordinate information is grouped for the jth after normalization；
Every five time points are one to coordinate information described in step 3 Jing Guo normalized AIS system acquisition sequentially in time
Group are as follows:
j∈[0,N-4]
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points,For jth
N-th of coordinate information in group, WSjCoordinate information is grouped for the jth after normalization；
By normalized high-precision GPS coordinate information, every five time points are one group sequentially in time described in step 3 are as follows:
j∈[0,N-4]
Wherein, time is acquisition initial time, and Δ t is acquisition interval, and N is acquisition points,For jth
N-th of coordinate information in group, WGjCoordinate information is grouped for the jth after normalization.
5. the ship track data fusion method according to claim 1 based on LSTM model, which is characterized in that step 4
Described in LSTM model be one have α input layer, β hidden layer, the structure of γ output layer, wherein LSTM learnt
Journey focuses on hidden layer using the distinctive cell unit of LSTM, which has input gate, forgets door and out gate；
Input gate indicates whether that allowing new data information to be added to currently hides in node layer, allows to input if value is 1, if
Value does not allow then for 0；
Forget door and indicates whether that the historical data for retaining current hiding node layer storage retains if value is that 1 i.e. door is opened, if
It is closed for 0 i.e. door, then empties the historical data that present node is stored；
Out gate indicates whether to export present node output valve to next layer, i.e., next hidden layer or output layer, if
It is opened for 1 i.e. door, then the output valve of present node will act on next layer, close if it is 0 i.e. door, present node output valve is not defeated
Out；
LSTM model described in step 4 is that crucial formula is as follows:
xtime+(i+1)*Δt=[WLj,WSj]
htime+i*Δt=[WLj,WSj,WGj]
ftime+(i+1)*Δt=σ (Wf·[Ctime+i*Δt,htime+i*Δt,xtime+(i+1)*Δt]+bf)
itime+(i+1)*Δt=σ (Wi·[Ctime+i*Δt,htime+i*Δt,xtime+(i+1)*Δt]+bi)
Otime+(i+1)*Δt=σ (Wo·[Ctime+(i+1)*Δt,htime+i*Δt,xtime+(i+1)*Δt]+bo)
Ctime+(i+1)*Δt=ftime+(i+1)*Δt*Ctime+i*Δt+itime+(i+1)*Δt
*tanh(Wc[htime+i*Δt,xtime+(i+1)*Δt]+bc)
htime+(i+1)*Δt=Otime+(i+1)*Δttanh(Ctime+(i+1)*Δt)
Wherein, its concrete meaning is as follows in above-mentioned formula, and f indicates that forgeing door, i expression input gate, O indicates that out gate, C indicate single
Member activation vector, h indicate to hide layer unit, WfTo forget the weight matrix between door and other input vectors, WiFor input gate with
Other receive the weight matrix between the vector of input data, WoFor the weight matrix between data before out gate and out gate, WcFor
Weight matrix between other vectors of unit activating vector, activation primitive are tanh, bf、bi、bo、bcRespectively forget door, input
The deviation of door, out gate, unit activating vector, different sampling moment indicate with subscript t, wherein htime+(i+1)*ΔtValue be
By the hidden layer h at last momenttime+i*ΔtWith new input xtime+(i+1)*ΔtAnd the oneself state of cell last moment
Ctime+i*Δt, controlled by door compound come in the h at the last one momenttime+(i+1)*ΔtAs export.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810537118.6A CN109001722B (en) | 2018-05-30 | 2018-05-30 | Ship track data fusion method based on LSTM model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810537118.6A CN109001722B (en) | 2018-05-30 | 2018-05-30 | Ship track data fusion method based on LSTM model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109001722A true CN109001722A (en) | 2018-12-14 |
| CN109001722B CN109001722B (en) | 2022-03-15 |
Family
ID=64573237
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810537118.6A Active CN109001722B (en) | 2018-05-30 | 2018-05-30 | Ship track data fusion method based on LSTM model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109001722B (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109711022A (en) * | 2018-12-17 | 2019-05-03 | 哈尔滨工程大学 | A kind of submarine anti-sinking system based on deep learning |
| CN109726355A (en) * | 2019-01-04 | 2019-05-07 | 重庆邮电大学 | A kind of ship track restorative procedure based on vector interpolation |
| CN109740742A (en) * | 2019-01-14 | 2019-05-10 | 哈尔滨工程大学 | A kind of method for tracking target based on LSTM neural network |
| CN110174690A (en) * | 2019-05-30 | 2019-08-27 | 杭州中科微电子有限公司 | A kind of satellite positioning method based on shot and long term memory network auxiliary |
| CN110290466A (en) * | 2019-06-14 | 2019-09-27 | 中国移动通信集团黑龙江有限公司 | Floor method of discrimination, device, equipment and computer storage medium |
| CN110519693A (en) * | 2019-09-29 | 2019-11-29 | 东北大学 | A kind of fusion and positioning method towards intelligent mobile terminal |
| CN110864440A (en) * | 2019-11-20 | 2020-03-06 | 珠海格力电器股份有限公司 | Air supply method, air supply device and air conditioner |
| CN110990504A (en) * | 2019-11-14 | 2020-04-10 | 中国船舶重工集团公司第七0七研究所 | Ship track compression method based on course and speed change rate |
| CN111257914A (en) * | 2020-01-14 | 2020-06-09 | 杭州电子科技大学 | Marine fishing boat track prediction method and system based on Beidou and AIS data fusion |
| CN111563072A (en) * | 2020-04-15 | 2020-08-21 | 交通运输部水运科学研究所 | AIS information-based ship real-time accurate position acquisition method |
| CN112268564A (en) * | 2020-12-25 | 2021-01-26 | 中国人民解放军国防科技大学 | Unmanned aerial vehicle landing space position and attitude end-to-end estimation method |
| CN112364119A (en) * | 2020-12-01 | 2021-02-12 | 国家海洋信息中心 | Ocean buoy track prediction method based on LSTM coding and decoding model |
| CN113516321A (en) * | 2021-09-14 | 2021-10-19 | 中国电子科技集团公司第十五研究所 | Model training method and device for predicting ship track |
| CN115346399A (en) * | 2022-07-23 | 2022-11-15 | 交通运输部规划研究院 | Bridge ship collision prevention early warning system based on phased array radar, AIS and LSTM network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103353756A (en) * | 2013-05-27 | 2013-10-16 | 武汉理工大学 | Method for monitoring underway ship in real time based on AIS and VTS information integration |
| CN105244020A (en) * | 2015-09-24 | 2016-01-13 | 百度在线网络技术（北京）有限公司 | Prosodic hierarchy model training method, text-to-speech method and text-to-speech device |
| US9760806B1 (en) * | 2016-05-11 | 2017-09-12 | TCL Research America Inc. | Method and system for vision-centric deep-learning-based road situation analysis |
| CN108022012A (en) * | 2017-12-01 | 2018-05-11 | 兰州大学 | Vehicle location Forecasting Methodology based on deep learning |
-
2018
- 2018-05-30 CN CN201810537118.6A patent/CN109001722B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103353756A (en) * | 2013-05-27 | 2013-10-16 | 武汉理工大学 | Method for monitoring underway ship in real time based on AIS and VTS information integration |
| CN105244020A (en) * | 2015-09-24 | 2016-01-13 | 百度在线网络技术（北京）有限公司 | Prosodic hierarchy model training method, text-to-speech method and text-to-speech device |
| US9760806B1 (en) * | 2016-05-11 | 2017-09-12 | TCL Research America Inc. | Method and system for vision-centric deep-learning-based road situation analysis |
| CN107368890A (en) * | 2016-05-11 | 2017-11-21 | Tcl集团股份有限公司 | A kind of road condition analyzing method and system based on deep learning centered on vision |
| CN108022012A (en) * | 2017-12-01 | 2018-05-11 | 兰州大学 | Vehicle location Forecasting Methodology based on deep learning |
Non-Patent Citations (2)
| Title |
|---|
| 权波等: "基于LSTM的船舶航迹预测模型", 《计算机科学》 * |
| 鹿强等: "海上目标多源轨迹数据关联综述", 《地球信息科学学报》 * |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109711022B (en) * | 2018-12-17 | 2022-11-18 | 哈尔滨工程大学 | Submarine anti-sinking system based on deep learning |
| CN109711022A (en) * | 2018-12-17 | 2019-05-03 | 哈尔滨工程大学 | A kind of submarine anti-sinking system based on deep learning |
| CN109726355A (en) * | 2019-01-04 | 2019-05-07 | 重庆邮电大学 | A kind of ship track restorative procedure based on vector interpolation |
| CN109726355B (en) * | 2019-01-04 | 2022-11-25 | 重庆邮电大学 | Ship track repairing method based on vector interpolation |
| CN109740742A (en) * | 2019-01-14 | 2019-05-10 | 哈尔滨工程大学 | A kind of method for tracking target based on LSTM neural network |
| CN110174690A (en) * | 2019-05-30 | 2019-08-27 | 杭州中科微电子有限公司 | A kind of satellite positioning method based on shot and long term memory network auxiliary |
| CN110290466A (en) * | 2019-06-14 | 2019-09-27 | 中国移动通信集团黑龙江有限公司 | Floor method of discrimination, device, equipment and computer storage medium |
| CN110519693A (en) * | 2019-09-29 | 2019-11-29 | 东北大学 | A kind of fusion and positioning method towards intelligent mobile terminal |
| CN110990504A (en) * | 2019-11-14 | 2020-04-10 | 中国船舶重工集团公司第七0七研究所 | Ship track compression method based on course and speed change rate |
| CN110864440A (en) * | 2019-11-20 | 2020-03-06 | 珠海格力电器股份有限公司 | Air supply method, air supply device and air conditioner |
| CN111257914A (en) * | 2020-01-14 | 2020-06-09 | 杭州电子科技大学 | Marine fishing boat track prediction method and system based on Beidou and AIS data fusion |
| CN111563072A (en) * | 2020-04-15 | 2020-08-21 | 交通运输部水运科学研究所 | AIS information-based ship real-time accurate position acquisition method |
| CN112364119A (en) * | 2020-12-01 | 2021-02-12 | 国家海洋信息中心 | Ocean buoy track prediction method based on LSTM coding and decoding model |
| CN112268564A (en) * | 2020-12-25 | 2021-01-26 | 中国人民解放军国防科技大学 | Unmanned aerial vehicle landing space position and attitude end-to-end estimation method |
| CN113516321A (en) * | 2021-09-14 | 2021-10-19 | 中国电子科技集团公司第十五研究所 | Model training method and device for predicting ship track |
| CN113516321B (en) * | 2021-09-14 | 2021-11-16 | 中国电子科技集团公司第十五研究所 | Model training method and device for predicting ship track |
| CN115346399A (en) * | 2022-07-23 | 2022-11-15 | 交通运输部规划研究院 | Bridge ship collision prevention early warning system based on phased array radar, AIS and LSTM network |
| CN115346399B (en) * | 2022-07-23 | 2024-01-19 | 交通运输部规划研究院 | Bridge ship collision prevention early warning system based on phased array radar, AIS and LSTM network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109001722B (en) | 2022-03-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109001722A (en) | A kind of ship track data fusion method based on LSTM model | |
| Bhatti et al. | Hostile control of ships via false GPS signals: Demonstration and detection | |
| US10424205B2 (en) | Auxiliary berthing method and system for vessel | |
| CN107577230A (en) | A kind of intelligent avoidance collision system towards unmanned boat | |
| CN102780523A (en) | Multi-satellite cooperative observation business scheduling method | |
| CN104049241B (en) | The spacing synchronization process of the double-base synthetic aperture radar that target location coordinate is unknown | |
| CN105116390A (en) | Marine radar calibration-oriented measured value and AIS truth value dot pair construction method | |
| CN104237862B (en) | Probability hypothesis density filter radar system error fusion estimation method based on ADS-B | |
| Jie et al. | A novel estimation algorithm for interpolating ship motion | |
| Wu et al. | A new multi-sensor fusion approach for integrated ship motion perception in inland waterways | |
| Czaplewski et al. | Improvement in accuracy of determining a vessel’s position with the use of neural networks ana robust m-estimation | |
| CN112433232B (en) | Search and rescue system and method based on Beidou high-precision positioning AIS terminal | |
| Wang et al. | A Jamming Aware Artificial Potential Field Method to Counter GPS Jamming for Unmanned Surface Ship Path Planning | |
| Ivanovsky et al. | Algorithm design for ship’s steering with specified limitations under various weather conditions | |
| Pan et al. | Research on Ship Arrival Law Based on Route Matching and Deep Learning | |
| Vicen-Bueno et al. | Live tasking/command and control (C2) of ISR unmanned underwater gliders from remote operational sites | |
| CN105676216A (en) | Low speed formation target tracking method based on formation pattern | |
| Liland | AIS aided multi hypothesis tracker-multi-frame multi-target tracking using radar and the automatic identification system | |
| Weintrit | Navigational Systems and Simulators | |
| Webb et al. | Applying prototype ship transit data to simulator validation | |
| Liu et al. | Towards intelligent navigation in future autonomous surface vessels: developments, challenges and strategies | |
| Bu | AIS-Data For Increased Insight Into Navigational Impacts Post Installation of Man-made Structures at Sea | |
| Lu et al. | BDS for Train Localization Performance Forecast in Railway Environments Using Machine Learning | |
| Praczyk | Application of neural networks and radar navigational aids of shore area to positioning | |
| Felski et al. | Monitoring of the movement of the objects on the Gdansk bay in order to recognize the characteristics of their main propulsion systems |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |