Systems, devices, and methods for predicting ship tracks

Abstract

Provided is a system and method for predicting a track of a ship. The prediction engine includes an advanced predictor configured to integrate ship track information with historical ship statistics and a land-water map, generate an optimized graph for a lowest-cost path from the ship track information, generate a time-likelihood map, and generate a total probability map from the optimized graph and the time-likelihood map to predict the track of the ship.

Description

SYSTEMS, DEVICES, AND METHODS FOR PREDICTING SHIP TRACKSTechnical Field

[0001] The following relates generally to maritime intelligence, and more particularly to systems and methods for predicting ship tracks.Introduction

[0002] It may be desirable to predict the track of a ship, including a future, past, or intermediate location of the ship. It may be desirable to know a track to a predicted point, including prediction uncertainty. Predictions may be used for planning, such as tasking satellite image, ship-to-track correlation, and / or user visualization.

[0003] However, given the diversity and complexity of ship behavior, and the computational challenge of extracting relevant historical behavior from complex and massive historical data, it is difficult to accurately predict ship tracks for a maritime vessel.

[0004] Accordingly, there is a need for an improved system and method for predicting ship tracks that overcomes at least some of the disadvantages of existing systems and methods.Summary

[0005] Provided is a prediction engine for predicting a track of a ship. The prediction engine includes an advanced predictor configured to integrate ship track information with historical ship statistics and a land-water map, generate an optimized graph for a lowest-cost path from the ship track information, generate a time-likelihood map, and generate a total probability map from the optimized graph and the timelikelihood map to predict the track of the ship.

[0006] The prediction engine may further include a ship statistic service configured to pre-compute historical statistics, provide the advanced predictor with statistical maps and land-water maps for a given area of interest, and provide distance to land for a given location of the ship.

[0007] The lowest-cost path may be generated using graph cost terms that include any one or more of ship count, heading statistics, land proximity, ship heading, track directness, and ship destination.

[0008] The optimized graph may be optimized by determining a best path to each pixel.

[0009] The advanced predictor may generate a predicted result. The predicted result may include any one or more of a predicted point, a predicted track, uncertainty polygons, and an uncertainty map.

[0010] The predicted result may include an uncertainty map. The uncertainty map includes a predicted location of the ship and a likelihood of the ship going to any other feasible location within the map.

[0011] The predicted result may include a video that shows predicted tracks of different vessels for varying times.

[0012] The predicted result may include predictions of any one or more of a future, past, intermediate location, a track to predicted point, and a prediction uncertainty.

[0013] The advanced predictor may receive a prediction request. The prediction request may include any one or more of ship track and prediction time.

[0014] The prediction engine may further include a great circle predictor that generates a prediction of a great circle of the ship, and a general predictor that determines whether to use the advanced predictor or the great circle predictor.

[0015] Provided is a method for predicting a track of a ship. The method includes integrating ship track information with historical ship statistics and a land-water map, generating an optimized graph for a lowest-cost path from the ship track information, generating a time-likelihood map, and generating a total probability map from the optimized graph and the time-likelihood map to predict the track of the ship.

[0016] The method may further include pre-computing historical statistics, providing statistical maps and land-water maps for a given area of interest, and providing distance to land for a given location of the ship.

[0017] The method may further include generating a prediction result.

[0018] The method may further include generating a prediction of a great circle of the ship, and determining whether to use the prediction of a great circle of the ship.

[0019] Other aspects and features will become apparent to those ordinarily skilled in the art, upon review of the following description of some exemplary embodiments.Brief Description of the Drawings

[0020] The drawings included herewith are for illustrating various examples of articles, methods, and apparatuses of the present specification. In the drawings:

[0021] Figure 1 is a block diagram of a system for predicting ship tracks, in accordance with an embodiment;

[0022] Figure 2 is a block diagram of a method for predicting ship tracks, in accordance with an embodiment;

[0023] Figures 3A, 3B, and 3C illustrate example maps of the method of Figure 2;

[0024] Figure 4 is a block diagram of a method for predicting ship tracks, in accordance with an embodiment;

[0025] Figure 5 is a block diagram of a method for ship statistics, in accordance with an embodiment;

[0026] Figure 6 is a graph of ship statistics that may be used in the system of Figure 1 and the methods of Figure 2, 4 and 5;

[0027] Figures 7A and 7B illustrate an example of forward prediction and backward prediction of a ship track, in accordance with an embodiment;

[0028] Figures 8A and 8B illustrate an example of uncertainty map without heading ambiguity and an uncertainty map with ambiguity, respectively, in accordance with an embodiment;

[0029] Figures 9A and 9B illustrate an example track prediction, in accordance with an embodiment; and

[0030] Figures 10A and 10B illustrate an example with known destinations, in accordance with an embodiment.Detailed Description

[0031] Various apparatuses or processes will be described below to provide an example of each claimed embodiment. No embodiment described below limits any claimed embodiment and any claimed embodiment may cover processes or apparatuses that differ from those described below. The claimed embodiments are not limited to apparatuses or processes having all of the features of any one apparatus or process described below or to features common to multiple or all of the apparatuses described below.

[0032] One or more systems described herein may be implemented in computer programs executing on programmable computers, each comprising at least one processor, a data storage system (including volatile and non-volatile memory and / or storage elements), at least one input device, and at least one output device. For example, and without limitation, the programmable computer may be a programmable logic unit, a mainframe computer, server, and personal computer, cloud-based program or system, laptop, personal data assistance, cellular telephone, smartphone, or tablet device.

[0033] Each program is preferably implemented in a high-level procedural or object-oriented programming and / or scripting language to communicate with a computer system. However, the programs can be implemented in assembly or machine language, if desired. In any case, the language may be a compiled or interpreted language. Each such computer program is preferably stored on a storage media or a device readable by a general or special purpose programmable computer for configuring and operating the computer when the storage media or device is read by the computer to perform the procedures described herein.

[0034] A description of an embodiment with several components in communication with each other does not imply that all such components are required. On the contrary, a variety of optional components are described to illustrate the wide variety of possible embodiments of the present invention.

[0035] Further, although process steps, method steps, algorithms or the like may be described (in the disclosure and I or in the claims) in a sequential order, such processes, methods and algorithms may be configured to work in alternate orders. In other words, any sequence or order of steps that may be described does not necessarily indicate a requirement that the steps be performed in that order. The steps of processes described herein may be performed in any order that is practical. Further, some steps may be performed simultaneously.

[0036] When a single device or article is described herein, it will be readily apparent that more than one device I article (whether or not they cooperate) may be used in place of a single device I article. Similarly, where more than one device or article is described herein (whether or not they cooperate), it will be readily apparent that a single device I article may be used in place of the more than one device or article.

[0037] Referring to Figure 1 , described therein is a system 100 for predicting ship tracks, in accordance with an embodiment. The system 100 includes a prediction engine 102 for predicting a track of a ship. The prediction engine 102 is provided on any one or more computer systems such as a computer server.

[0038] The system 100 includes an external computer system 104, which may be a separate computer system from the computer system of the prediction engine 102.

[0039] The system 100 includes a ship 106 that is situated and moving within an environment 108. The ship 106 may be a maritime vessel. When ships are navigating around land, or storms, or other marine traffic, or approaching port, the path of the ship may be difficult to predict. The ship track includes a ship’s location at a particular time and may include self-reported ship information including, kinematic values, uncertainties, data sources, and quality information.

[0040] The system 100 includes sensors 110 that sense the ship 106 and the environment 108 as the ship 106 navigates within the environment 106. The sensors 110 may provide image data to the external system 104 or the prediction engine 102.

[0041] The prediction engine 102 includes the ship statistic service 112. The ship statistic service 112 receives historical data 114. The historical data 114 includes shipstatistics and land-water maps. The historical ship data includes historical automatic identification system (AIS) data, which may be stored as vector data. The historical data 114 may be stored as images and / or rasters. The historical data as represented by 114 is vector AIS data. The pre-processing step converts that AIS data into statistical raster maps and generic statistics. The raster includes a matrix of cells or pixels organized into rows and columns or a grid where each cell contains a value representing information.

[0042] The external computer system 104 includes a user interface 116 where a user inputs a prediction request 118. The prediction request 118 includes any one or more of ship track and prediction time. The external system 104 sends the prediction request 118 to the prediction engine 102. The prediction engine 102 generates a prediction result 120. The prediction result 120 includes any one or more of a predicted point, a predicted track, uncertainty polygons, and an uncertainty map. Uncertainty polygons include contours of the uncertainty map at specific uncertainty levels.

[0043] The prediction engine 102 receives the prediction request 118 including a vector track description of a current ship track from the external system 104 to predict a ship track that is optimized against multiple criteria. The ship statistic service 112 may provide the distance the ship is from land 128.

[0044] The prediction engine 102 includes a general predictor 122. The general predictor 122 interprets the prediction request 118. The general predictor 122 extracts data from the prediction request 118 for the track prediction. The general predictor 122 queries for prediction with the extracted data.

[0045] The prediction engine 102 includes at least one predictor. The prediction engine 102 includes a great circle predictor 124. The prediction engine 102 includes an advanced predictor 126.

[0046] The general predictor 122 determines which of the great circle predictor 124 or the advanced predictor 126 to use. The ship statistic service 112 calculates the distance to land 128 for any given ship location. The general predictor 122 uses distance- to-land data 128 from the ship statistics service 112 to determine which predictor (124, 126) to use.

[0047] The advanced predictor 126 receives a prediction query 130 from the general predictor 122. The advanced predictor 126 requests (at 132) statistical maps 134 from the ship statistics service 112. The advanced predictor 126 receives the statistical maps 134 from the ship statistics service 112.

[0048] The advanced predictor 126 generates the prediction result 120, utilizing statistical information within an optimization process. The advanced predictor sends the prediction result 120 to the general predictor 122. The general predictor 122 returns the prediction result 120 to the external system 104.

[0049] The ship statistics service 112 compresses the historical data 114 and sends the compressed pre-computed statistics 134 to the advanced predictor 126.

[0050] The ship statistic service 104 also provides a statistics map 136 to the external system 104.

[0051] The ship statistic service 112 uses the historical ship statistics to generate pre-computed ship statistics 134. The pre-computed ship statistics 134 may be worldwide and enable fast predictions. The pre-computed ship statistics may include pre-generated rasters to enable fast predictions. The pre-generated rasters include land maps that allow fast land lookups, that may be on the same grid as the statistics. The land map may include land / water ratio for mixed cells. The pre-generated rasters allow the prediction engine 102 to quickly retrieve and utilize information needed for the prediction, enabling fast computations rather than spending computer resources computing statistics and transforming data at request time.

[0052] The advanced predictor 126 integrates ship track information with historical ship statistics and the land-water map 134. The advanced predictor 126 optimizes a best path to each pixel using any one or more of travel distance, statistical likelihood, land avoidance, and initial bearing. The advanced predictor 126 generates a cost map from the best path to each pixel. The advanced predictor 126 generates a time-aware probability map to predict the track of the ship.

[0053] The advanced predictor 126 may include a graph optimization algorithm to find a track optimized for multiple criteria. In an example, the optimization algorithm is aDijkstra algorithm. The optimization criteria includes factors for track directness, land avoidance, adherence to statistics, and maintenance of overall direction.

[0054] The great circle predictor 124 generates a great circle prediction 138 of a great circle of the ship. When travelling on the open ocean, a ship will typically follow a great circle trajectory which is relatively simple to predict.

[0055] The prediction engine 102 enforces land avoidance, while utilizing multiple information sources 112 to guide the general predictor 122 to a realistic ship track prediction 120. The advanced predictor 126 and the general predictor 122 are computationally fast allowing for interactive predictions.

[0056] The prediction result 120 includes an uncertainty map that demonstrates the likelihood of the ship going to other locations besides the predicted location.

[0057] The prediction result 120 includes a ship track that may be realistically predicted to provide comprehensive maritime awareness and displaying expected ship motion to users. The prediction result 120 indicates uncertainty, so that a user can understand how confident the model is in its prediction.

[0058] The prediction result 120 may include a video that shows predicted tracks of different vessels for varying times (e.g., 10 minutes to 12 hours). The predicted tracks of the vessels may accurately follow known routes which vessels typically take. The prediction result 120 may include uncertainty polygons. The uncertainty polygons may represent 25%, 50%, and 75% confidence regions of where the vessel may actually be at any given prediction time. The predicted tracks and uncertainty polygons accurately avoid land.

[0059] The prediction engine 102 advantageously supplements current track information with pre-computed ship statistics (from advanced predictor 126) and the land map 134. The prediction engine 102 is optimized using a fast graph-optimization strategy to generate realistic and fast predictions. The prediction engine 102 makes a globally optimized prediction given the prediction window, rather than just making a sequence of locally-informed steps which could result in illogical global predictions.

[0060] The inputs may also include an AIS global positioning system (GPS) track or a single satellite image of a ship, such as a synthetic aperture radar (SAR) image with a ship detection. The steps to convert satellite imagery or other sensor data into “detections” is outside of the predictor engine 102. The input to the predictor engine 102 includes vector information representing a ship track. The ship track may be entirely based on a single satellite image, where that information is represented as a track. It is the task of the prediction engine to predict the most likely track given the input track.

[0061] In an embodiment, the ship statistics service 112 builds worldwide statistical maps from the database of multi-year AIS data. Building these maps make take weeks to months of compute time. The ship statistics service 112 optimally fits ship speed, heading, and other factors to low-dimension distributions parameterized by the ships kinematics, location, and prediction time window.

[0062] In an embodiment, the prediction engine 102 may operate without the statistical raster maps (which are a compression of the historical data). Once the statistical raster maps are computed (prior to live usage), the historical database may be disconnected. The prediction engine 102 still uses the land-map and distance to land data 128. In an embodiment, the general predictor 122 queries the statistics service for distance-to-land information.

[0063] Referring to Figure 2, described therein is a method 200 for predicting ship tracks, in accordance with an embodiment. Figures 3A and 3B illustrate intermediate calculations 300, 302, and Figure 3C illustrates an example output 304, of the method 200.

[0064] At 202, historical ship statistics are fetched. The historical ship statistics 306 include location-dependent and generic ship behavior. The ship track information is integrated with the historical ship statistics 306 and land-water maps 308.

[0065] At 204, the best path to each pixel is optimized. The best path is determined from an input track 310. The optimization criteria includes any one or more of travel distance, statistical likelihood, land avoidance, and initial bearing. Directional statistics are used to navigate traffic separation. A cost map 302 is generated.

[0066] At 206, a time-aware probability map 304 is generated to determine a predicted track 316 of a ship. The method 200 may generate both the predicted track 316 and an uncertainty map 304. The uncertainty map 304 may include contours 318a, 318, indicting the likelihood of the ship tracking in that location.

[0067] The method 200 may generate an output that includes predictions of any one or more of a future, past, or intermediate location, a track to predicted point, and a prediction uncertainty. The predictions may be used for any one or more of planning (e.g., tasking satellite image, ship-to-track correlation, and user visualization).

[0068] Referring to Figure 4, described therein is a method 400 for predicting ship tracks, in accordance with an embodiment. The method 400 may be performed, for example by the advanced predictor 126 of Figure 1.

[0069] Initially, input information is received. The input information includes a prediction request that includes the known ship track and prediction time.

[0070] The method 400 may predict forward in time (where the track is expected to go after its latest known location), backward in time (the track the ship was expected to follow to arrive at its known locations), and / or interpolation between known track locations. The prediction time, with respect to the times of the known input track, indicates whether forward, backward, or interpolation will satisfy the prediction request.

[0071] At 402, the Area of Interest (AOI) is defined for given prediction time window. The user may create the AOI that is large enough given the input track including ship speed and prediction time.

[0072] At 404, the land and statistical maps for the AOI are received. The land map indicates which pixels are land, water, or a mixture of land and water. The statistical maps include ship counts for each pixel in the AOI. The statistical maps include ship heading statistics for each pixel in the AOI, that indicate the probability of a ship heading in any given direction for that pixel.

[0073] At 406, a mask of feasible pixels are determined.

[0074] At 408, a heading map is generated emphasizing the most likely directions the ship will go given the ship’s current heading. The heading map is calculated fromgeneric heading statistics from the ship statistics service. The heading map causes the prediction to keep the ship’s heading unless statistics or land indicate otherwise.

[0075] At 410, a single graph with cost per edge is generated from a fusing of statistics, land, and heading maps. The fused map is a weighted average of count statistics, heading statistics, course map, and land map. Partial land pixels may be discouraged. The fused map illustrates the desirability of each pixel in the image.

[0076] The graph defines how the ship might move from one pixel to another pixel. Each edge has an edge cost. The system generates the pixel graph based on the edge cost. The edge costs are based on physical distance and fused map value. Edges over infeasible pixels are excluded from the graph.

[0077] At 412, the graph is optimized to find the lowest-cost path to each pixel on the map, which is termed a cost map. The lowest cost path to each pixel may be found using the Dijkstra algorithm. The Dijkstra algorithm finds the shortest path between a given node (which is called the "source node") and all other nodes in a graph. The Dijkstra algorithm uses the costs of the edges to find the path that minimizes the total cost between the source node and all other nodes. The cost map is not time-aware.

[0078] At 414, a time-likelihood map is generated. The time-likelihood map applies a distance likelihood distribution to the cost-optimized distances from the optimized graph. The distance likelihood distribution comes from the generic statistics of ship statistics service. The time-likelihood map indicates the likelihood of the ship being at any given pixel in the map given the prediction time and the ship’s initial velocity.

[0079] At 416, a total probability map is generated by overlaying the cost map (representing spatial likelihood) with the time-likelihood map. The total probability map is a predicted result.

[0080] The prediction result may include a predicted point that is a point of highest probability. The prediction result may include a predicted track that is extracted from the optimized graph. The prediction results may include uncertainty polygons that show likelihood that a ship is within the region, which are contours of the total probability map. Advantageously, instead of calculating spatio-temporal likelihood simultaneously, themethod 400 optimizes for spatial likelihood at 412, then overlays the temporal likelihood at 414 together at 416. The method 400 advantageously decouples the spatial likelihood and time likelihood to enable predictions that can be computed faster (e.g., within seconds).

[0081] Referring to Figure 5, described therein is a method 500 for ship statistic generation, in accordance with an embodiment. The method 500 may be performed, for example, by the ship statistic service 112 as described with reference to Figure 1 .

[0082] At 502, historical statistics are received. The historical data includes historical AIS data. The AIS data may be represented as ship tracks. The historical statistics are computed from a massive AIS database and may include worldwide data. The historical statistics include generic statistics. The historical statistics include raster statistics. The historical statistics are stored as raster maps.

[0083] At 504, the historical statistics are pre-computed. The generic statistics data and the historical statistic data are pre-processed.

[0084] At 504, the generic statistics data is pre-processed. Parameters in fewparameter distribution models are optimized to best represent the given AIS tracks. For example, the mean and standard deviation of the Normal distribution may be defined as a function of ship speed, ship type, distance from land, and time. The parameterized distributions represent ship behaviors such as likelihood to change heading and likelihood of track length. Optimized parameters are saved for use in prediction.

[0085] At 504, the raster statistics data is pre-processed. AIS tracks are overlaid onto a worldwide raster grid and each track contributes to the density, expected heading, and expected speed statistics for each cell the track crosses. The statistical raster maps are saved.

[0086] The method 500 performed by a computer system performs significant computation. The method 500 advantageously computes through very large data and very time consuming operations as a pre-processing stage, and saves the statistics in a compressed and accessible format that can be rapidly accessed upon request. The method 500 advantageously utilizes parameterized distributions and statistical rastermaps to compress terabytes of data into simplified data that can be quickly utilized for the purpose of ship prediction.

[0087] The optimized parameters from the generic statistics are stored in computer memory. The statistical raster maps are stored in memory. The worldwide land and distance to land rasters are stored in computer memory.

[0088] At 506, statistical maps and land / water maps are provided for a given AOI. Distance to land for a given location is provided. The ship statistics service delivers statistics. The ship statistics service receives a request from a component, such as a ship track predictor. The ship statistics service fetches the appropriate data from the stored data. The desired data is delivered to the component (e.g., predictor).

[0089] The method 500 is constructed to enable fast (interactive) predictions because the method 500 pre-processes the historical data at 504, rather than when the data is requested from the component. Pre-processing the generic statistics data and raster statistics data may be computationally slow. Since the pre-processing is performed before the component requests the data, the data is able to be delivered in real-time. The pre-computed statistics are saved, they can be quickly utilized whenever the predictor requests statistics.

[0090] When generating the pre-compute statistics, the method 500 determines the correct statistics to extract as well as what ship categories to consider, while making this computationally feasible at a worldwide scale.

[0091] The method 500 develops statistical maps and models of typical ship behavior.

[0092] The historical ship statistics may be built from AIS data. Historical statistics are a product derived from historical data. Ship statistics may be derived per ship group. The ship statistics may be worldwide (geo-aware) and / or generic (geo-agnostic). Ship groups may include all ships. Ship groups may be defined according to ship lengths. Ship groups may defined according to ship types; for example, including cargo, tanker, passenger, fishing, and other.

[0093] The historical ship statistics may include generic statistics of a ship. The general statistics use functional and parameterized distributions to represent the data. The generic statistic characterizes ship behavior agnostic of geolocation. The generic statistic may be generated from a large set of ship tracks with normalized location and heading.

[0094] The historical ship statistics may include worldwide statistics and define basic statistics on a raster map of the entire world. The worldwide statistics include any one or more of ship count, heading, and average speed. The historical ship statistics may include statistics evolving over time. The historical ship statistics may include statistics varying by initial velocity. The historical ship statistics may include statistics varying ship type. The historical ship statistics may include statistics varying by distance from land.

[0095] Figure 6 illustrates a graph 600 of track distance statistics with distribution 602 against final distance 604, in accordance with an embodiment. The curves 606a, 608a, 610a, 612a, 614a, 616a, are shown in the plot in solid lines. The different curves 606a, 608a, 610a, 612a, 614a, 616a, indicate different ship velocity ranges, and the curve data represents the distribution of how ships with that given velocity travel over the given period of time (3 hours in this example).

[0096] Normal distributions 606b, 608b, 610b, 612b, 614b, 616b in dotted lines, for each of the track curves 606a, 608a, 610a, 612a, 614a, 616a (respectfully) are optimized to fit the data. The optimization has the best-fit distributions follow a smooth trajectory over time, over ship velocities, and over distance-from-land. While the data distribution curves (606a, 608a, 610a, 612a, 614a, 616a) provide useful intuition, the parameterized distributions provide a compact function that represents the distribution of ship behaviors given the known ship location, type, velocity, and prediction time. Figure 6 is an example of a generic statistics distribution for ship travel distance, but similar distributions are also defined for ship heading change.

[0097] Figures 7A and 7B illustrate an example of forward prediction 700 and backward prediction 702. The system can predict forward track 700 and / or backward track 702 in time from the same initial track 704. The forward prediction 700 shows thepredicted track 706. The backward prediction 702 shows the possible backward track 708 and potential areas of origin 710.

[0098] Figures 8A and 8B illustrate an example showing the resulting uncertainty map when the input track does not have a heading ambiguity 800 and the uncertainty map when the input track does have a heading ambiguity 802. A ship track extracted from an optical or SAR image will often have a 180° ambiguity for the ships current heading. Figures 8A and 8B reflect the uncertainty map 800, 802 and contours.

[0099] Figures 9A and 9B illustrate an example in the English Channel 900, 902. Heading statistics cause the prediction to select the correct shipping lane 904.

[0100] Figures 10A and 10B illustrate an example with known destinations 1000, 1002 using known destinations guide the prediction. In Figure 10A the destination 1004 cannot be reached within the prediction time. In Figure 10B the destination 1006 can be reached within the prediction time.

[0101] While the above description provides examples of one or more apparatus, methods, or systems, it will be appreciated that other apparatus, methods, or systems may be within the scope of the claims as interpreted by one of skill in the art.

Claims

Claims:

1. A prediction engine for predicting a track of a ship, the prediction engine comprising: an advanced predictor configured to: integrate ship track information with historical ship statistics and a landwater map; generate an optimized graph for a lowest-cost path from the ship track information; generate a time-likelihood map; and generate a total probability map from the optimized graph and the timelikelihood map to predict the track of the ship.

2. The prediction engine of claim 1 further comprising a ship statistic service configured to: pre-compute historical statistics; provide the advanced predictor with statistical maps and land-water maps for a given area of interest; and provide distance to land for a given location of the ship.

3. The prediction engine of claim 1, wherein the lowest-cost path is generated using graph cost terms that include any one or more of ship count, heading statistics, land proximity, ship heading, track directness, and ship destination.

4. The prediction engine of claim 1 , wherein the optimized graph is optimized by determining a best path to each pixel.

5. The prediction engine of claim 1 , wherein the advanced predictor generates a predicted result, and wherein the predicted result includes any one or more of a predicted point, a predicted track, uncertainty polygons, and an uncertainty map.

6. The prediction engine of claim 1 , wherein the advanced predictor generates a predicted result, and wherein the predicted result includes an uncertainty map, wherein the uncertainty map includes a predicted location of the ship and a likelihood of the ship going to any other feasible location within the map.

7. The prediction engine of claim 1 , wherein the advanced predictor generates a predicted result, and wherein the predicted result includes a video that shows predicted tracks of different vessels for varying times.

8. The prediction engine of claim 1 , wherein the advanced predictor generates a predicted result, and wherein the predicted result includes predictions of any one or more of a future, past, intermediate location, a track to predicted point, and a prediction uncertainty.

9. The prediction engine of claim 1 , wherein the advanced predictor receives a prediction request, wherein the prediction request includes any one or more of ship track and prediction time.

10. The prediction engine of claim 1 further comprising: a great circle predictor that generates a prediction of a great circle of the ship; and a general predictor that determines whether to use the advanced predictor or the great circle predictor.

11. A method for predicting a track of a ship, the method comprising: integrating ship track information with historical ship statistics and a land-water map; generating an optimized graph for a lowest-cost path from the ship track information; generating a time-likelihood map; and generating a total probability map from the optimized graph and the time-likelihood map to predict the track of the ship.

12. The method of claim 11 further comprising: pre-computing historical statistics; providing statistical maps and land-water maps for a given area of interest; and providing distance to land for a given location of the ship.

13. The method claim 12, wherein the lowest-cost path is generated using graph cost terms that include any one or more of ship count, heading statistics, land proximity, ship heading, track directness, and ship destination.

14. The method claim 12, wherein the optimized graph is optimized by determining a best path to each pixel.

15. The method of claim 12 further comprising: generating a prediction result, wherein the prediction result includes any one or more of a predicted point, a predicted track, an uncertainty polygon, and an uncertainty map.

16. The method of claim 15, wherein the predicted result includes an uncertainty map, wherein the uncertainty map includes a predicted location of the ship and a likelihood of the ship going to any other feasible location around the ship.

17. The method of claim 15, wherein the prediction result includes a video that shows predicted tracks of different vessels for varying times.

18. The method of claim 13, wherein the predicted result includes predictions of any one or more of a future, past, intermediate location, a track to predicted point, and a prediction uncertainty.

19. The method of claim 12 further comprising: receiving a prediction request, wherein the prediction request includes any one or more of ship track and prediction time.

20. The method of claim 12 further comprising: generating a prediction of a great circle of the ship; and determining whether to use the prediction of a great circle of the ship.

Images