Computer system and method for mapping vessel track behaviour

Abstract

A system and method for mapping vessel track behaviour is provided herein. The system includes an activity detector for detecting vessel activity of a track. The system includes a modeler and trainer for modeling and classifying vessel activity of the track. The modeler and trainer is configured to set up configuration parameters for a model, and train the model to determine state transitions. The system includes a track activity segementer for segmenting vessel activity of the track. The trained model is used to infer a state transition sequence for the track. The track activity segementer is configured to add the inferred state transition sequence to the track, determine characteristic vector per segments by same state per track, and map characteristic vectors of segments into activities with a trained classifier.

Description

COMPUTER SYSTEM AND METHOD FOR MAPPING VESSEL TRACK BEHAVIOURTechnical Field

[0001] The following relates generally to vessel behaviour analytics, and more particularly to systems and methods for mapping vessel track behaviour.Introduction

[0002] Vessel track systems, including those to the same inventors, are described in A. Y. Shahir, T. Charalampous, M. A. Tayebi, U. Glasser, and H. Wehn, “TripTracker: Unsupervised Learning of Fishing Vessel Routine Activity Patterns,” Proc. - 2021 IEEE Int. Conf. Big Data, Big Data 2021 , pp. 1928-1939, 2021”, hereinafter “TripTracker”.

[0003] TripTracker describes a framework to monitor routine fishing activity patterns and identify vessel trip types from automatic identification system (AIS) vessel traffic data. The framework includes inputting raw AIS fishing traffic data, and outputting segmented end-to-end AIS fishing trips into routine vessel activities.

[0004] However, it may be desirable to reduce the number of AIS input features needed, and optimize model training for operationalization, deployment, and continuous monitoring.

[0005] Accordingly, there is a need for an improved system and method that overcomes at least some of the disadvantages of existing systems and methods.Summary

[0006] Provided is a system for mapping vessel track behaviour. The system includes an activity detector for detecting vessel activity within a track, a modeler and trainer for modeling and classifying vessel activity of the track. The modeler and trainer is configured to set up configuration parameters for a model, and train the model to determine state transitions. The system includes a track activity segementer for segmenting vessel activity of the track. The trained model is used to infer a state transition sequence for the track. The track activity segementer is configured to add the inferred state transition sequence to the track, determine characteristic vector per segments bysame state per track, and map characteristic vectors of segments into activities with a trained classifier.

[0007] The model may be a hidden Markov model (HMM).

[0008] The configuration parameters may include any one or more of hidden states, speed over ground (SOG) and course over ground (COG) features as observations.

[0009] The activity detector may be configured to determine a timer interval for micro-activities, shred pre-processed tracks into micro-activity tables, and determine a characteristic vector per micro-activity.

[0010] The activity detector may be configured to cluster a subset of characteristic vectors to identify distinct activities, and validate the cluster analysis by running a clustering algorithm on different sample data.

[0011] The activity detector may be configured to determine characteristic vectors per micro-activity, train a classifier using micro-activity characteristic vectors and cluster identifiers, and validate classifier performance using validation techniques.

[0012] The system may further include a track preprocessor for pre-processing the track. The track preprocessor may be configured to validate the format for the track and the features of the track and interpolate the track features to one position per a given time period.

[0013] The track may be an AIS track. The AIS track may include any one or more of latitude, longitude, time stamp, speed over ground (SOG), course over ground (COG), and maritime mobile service identity (MMSI).

[0014] The system may further include track post processor for post processing the track. The track post processor may be configured to set up visualization configuration parameters, and apply desired configuration to the track.

[0015] The activity may include any one or more of maneuvering, fishing, stationary, and steaming.

[0016] Provided is a method for mapping vessel track behaviour. The method includes detecting vessel activity of a track, modeling and classifying vessel activity of the track including setting up configuration parameters for a model and training the model to determine state transitions. The method includes segmenting vessel activity of the track. The trained model is used to infer a state transition sequence for the track. Segmenting vessel activity of the track includes adding the inferred state transition sequence to the track, determining characteristic vector per segments by same state per track, and mapping characteristic vectors of segments into activities.

[0017] The method may further include determining a timer interval for microactivities, shredding pre-processed tracks into micro-activity tables, and determining a characteristic vector per micro-activity.

[0018] The method may further include clustering a subset of characteristic vectors to identify distinct activities, and validating the cluster analysis by running a clustering algorithm on different sample data.

[0019] The method may further include determining characteristic vectors per micro-activity, training a classifier using micro-activity characteristic vectors and cluster identifiers, and validating classifier performance using validation techniques.

[0020] The method may further include validating the format for the track and the features of the track, and interpolating the track features to one position per a given time period.

[0021] The method may further include setting up visualization configuration parameters, and applying desired configuration to the track.

[0022] 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

[0023] The drawings included herewith are for illustrating various examples of articles, methods, and apparatuses of the present specification. In the drawings:

[0024] Figure 1 is a block diagram of a system for mapping vessel track behaviour, according to an embodiment;

[0025] Figure 2 is a flow chart of a method for preprocessing tracks, according to an embodiment;

[0026] Figure 3 is a flow chart of a method for activity detection, according to an embodiment;

[0027] Figure 4 is a flow chart of a method for activity detection, according to an embodiment;

[0028] Figure 5 is a flow chart of a method for modeling and training, according to an embodiment;

[0029] Figure 6 is a flow chart of a method for activity segmentation, according to an embodiment;

[0030] Figure 7 is a flow chart of a method for track post processing, according to an embodiment; and

[0031] Figure 8 is an image of an AIS track, according to an embodiment.Detailed Description

[0032] 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.

[0033] 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, amainframe computer, server, and personal computer, cloud-based program or system, laptop, personal data assistance, cellular telephone, smartphone, or tablet device.

[0034] 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.

[0035] 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.

[0036] 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.

[0037] 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.

[0038] Referring now to Figure 1 , shown therein is block diagram of a system 100 for mapping vessel track behavior, according to an embodiment. The system 100 automates and enhances the processes of vessel activity identification and detection.

[0039] The system 100 may be a vessel track behaviour mapper. The system 100 may provide vessel behaviour analytics for maritime domain awareness. The system 100 uses unsupervised learning algorithms to tackle complex technological computational tasks including data mining and / or time series segmentation. While the system 100 may be used in maritime applications, the system 100 may be used in behavioural applications more broadly, including fishing vessels, cargo vessels, military vessels, to detect different types of activities including surfacing and detecting. The system 100 may provide a methodology for a behavioural analytic framework.

[0040] The system 100 detects, identifies, and segments routine vessel activity patterns from automatic identification system (AIS) vessel traffic data. The vessel may be a maritime vessel including fishing-vessels AIS traffic data, and different AIS vessel traffic data and generalized tracks. The system 100 may provide a more efficient engine towards operationalization and / or deployment of this technology.

[0041] The system 100 may be applied to different AIS vessel traffic data and generalized tracks. Generalized tracks are tracks including AIS position reports and position reports from other sensors e.g. SAR, Electro-Optical, VIIRS. The AIS position report includes a location with latitude, longitude, and time stamp. The system 100 represents a more efficient engine towards operationalization and / or deployment of behaviour mapping.

[0042] The AIS data may include maritime traffic captured from a satellite. The satellite may take periodic images (e.g., at 4 days, 28 days). The AIS data may have images and include a detection latitude, longitude, and time.

[0043] The system 100 may combine a plurality of measurements from a plurality of satellites. The system 100 may use other types of measurements, from different satellites, to form a more complete picture of the vessel’s behaviour.

[0044] Datasets may include fishing traffic datasets including AIS fishing vessel traffic data, including commercial fishing vessel and fish processing vessels. Datasets may include endpoint datasets including identified locations where vessels congregate. Datasets may include fishing effort datasets including information about fishing effort and vessel presence of different geographical locations in a grid structure, and any one ormore of information about vessel type, number of vessels, hours of vessels, and hours of fishing activity binned into grid cells in a particular amount of degrees in latitude and longitude. Datasets may include standard AIS categories of data such as static information, navigational information, and voyage-specific information.

[0045] Datasets may include both end-to-end trip and partial trip extraction workflow. The data inputs include any one or more of raw AIS trajectories from fishing vessels during a given time interval and an area of interest (AOI), and reliable information about maritime anchorages (ports or harbours), also called endpoints, existing in the AOI. The system 100 includes a track preprocessor 102 for pre-processing a track. The track preprocessor may represent features of the track. The system 100 receives raw AIS tracks, and outputs segmented AIS tracks into routine fishing-vessel activities. Given an AIS track, the system 100 divides the segmented AIS tracks into a sequence of segments in order to reveal routine fishing vessel activities performed by the vessel along the track.

[0046] The outputs include end-to-end and partial fishing tracks. End-to-end tracks are also called end-to-end trips. That is, AIS fishing tracks that start at one confirmed maritime anchorage and end at another confirmed maritime anchorage. Partial tracks do not need to start or end at a confirmed anchorage point. Output tracks can be both AIS tracks or generalized tracks, according to the input.

[0047] The system 100 includes an activity detector 104 for detecting the activity of the track.

[0048] The system 100 includes a modeler and trainer 106 for modeling and training on the data.

[0049] The system 100 includes a track activity segementer 108 for segmenting vessel activities of a track.

[0050] The system includes a track post processor 110 for post processing the track.

[0051] The system 100 reduces the number of AIS input features needed, optimizing model training for operationalization, deployment, and continuous monitoring. The optimization in model training replaces a plurality (e.g., 30+) of trained models (e.g.,Hidden Markov Models (HMMs)) with a single model without compromising output quality. As a result, the system 100 yields faster computational performance than conventional systems (e.g., TripTracker) and may be easier to maintain in operation. More specifically, having only one model eliminates a step called “HMM Picker” in TripTracker. This step involved selecting the appropriate HMM (from among more than 30 HMMs) for segmenting a new trip.

[0052] In addition, the pre processor 102, and the post processor 110 allow the system 100 to handle more general AIS tracks as inputs / outputs. As a result, the preprocessor 102 can handle both AIS tracks or generalized tracks.

[0053] The system 100 may provide advanced analytics for maritime intelligence to realize insights extraction. The system 100 integrates unsupervised learning techniques for tackling the challenges of identification and segmentation of vessel activities in AIS tracks.

[0054] The system 100 may identify and segment other types of vessel behaviours / activities from diverse AIS maritime traffic datasets. The system 100 may provide more advanced vessel behaviour analytics that build upon the currently identified behaviours.

[0055] Referring now to Figure 2, shown therein is flow chart of a method 200 for preprocessing tracks, according to an embodiment. The method 200 may be performed by the track preprocessor of 102 of Figure 1 .

[0056] At 202, a track is received. The track may be an AIS track or a generalized track. The track may include any one or more of latitude, longitude, time stamp (e.g., UTC), speed over ground (SOG), course over ground (COG), and maritime mobile service identity (MMSI).

[0057] At 204, the track is pre-processed. Pre-processing of the track may include validating the format for the track and the features of the track. Pre-processing may include recalculating SOG and COG. Pre-processing may include calculating derived feature of a delta COG. The track is better to use after pre-processing the track is better.For example, the pre-processing removes noise in the SOG and COG. The preprocessing includes re-calculations of features (e.g., SOG, COG).

[0058] The pre-processing includes recalculating features of interest for each trip based on the latitude and longitude values. The directions of a vessel may be initially very inaccurate and inconsistent. The pre-processing includes using the Azimuth method to determine the direction between two consecutive points, taking into account the spherical shape of the Earth. The pre-processing includes calculating the difference in direction, known as delta Course Over Ground (COG).

[0059] The Speed Over Ground (SOG) can be determined based on the distance a vessel has traveled (using latitude and longitude) and the time it took to travel that distance. In some cases, more than 95% of the SOG values in the data may be close to the recalculated values. The pre-processing may intentionally not include this computationally intensive step in the preprocessing, to save computational resources.

[0060] At 206, the track is interpolated. Iterpolating the track may include interpolating the track features to one position per a given time period (e.g., one minute). Interpolation the track may include using a great circle arc interpolation between two points of the track.

[0061] An input satellite image includes the observation of some vessels or ships. The location of the vessel is interpolated based on candidate vessels known from existing AIS tracks to the acquisition time of the image. These inferred candidate vessels are used for matching with the observed ships.

[0062] At 208, the track is saved as the pre-processed track 210. The resulting pre- processed track may be written to a comma separated value (CSV) file.

[0063] The method 200 is illustrated for a single track however, the method 200 may input a single or multiple tracks. The output of the method corresponds to a given input (i.e. , one or multiple tracks).

[0064] Referring now to Figure 3, shown therein is flow chart of a method 300 for activity detection, according to an embodiment. The method 300 may be performed by the activity detector 104 of Figure 1 .

[0065] The method 300 may further include different features such as removing a distance to port. The input features input features may change depending on the activities that are targeted depending on the context of the application.

[0066] At 302, end-to-end tacks are stored in a database.

[0067] At 304, a timer interval (window) is determined for micro-activities. The method 300 may use 30 minutes as determined via data analysis.

[0068] At 306, the pre-processed tracks (e.g., 210), are shred into micro-activity tables.

[0069] At 308, a characteristic vector per micro-activity is determined. The arithmetic average of selected features are taken to create a single vector. The single vector may include the average SOG, the average delta COG, and the average distance to port.

[0070] At 310, a subset of characteristic vectors are clustered to identify distinct activities. The subset of characteristic vectors may be 20% of the characteristic vectors. The clustering may be performed by a clustering engine. The clustering engine may include a HDBSCAN algorithm. The HDBSCAN algorithm extends a Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm by converting the DBSCAN algorithm into a hierarchical clustering algorithm, and then using a technique to extract a flat clustering based in the stability of clusters.

[0071] At 312, the cluster analysis is validated by running the clustering algorithm on different sample data. The clustering algorithm may run on, for example, 50 different sample data.

[0072] At 314, the subset of micro-activities with assigned cluster identifiers is stored.

[0073] At 316, an analyst interprets clusters into activities by using domain knowledge and micro-activity features.

[0074] At 318, routine vessel activities are stored. The routine vessel activities include fishing, maneuvering, steaming, and stationary.

[0075] Referring now to Figure 4, shown therein is flow chart of a method 400 for activity detection, according to an embodiment. The method 400 may be performed by the activity detector 104 of Figure 1 .

[0076] At 402, micro-activities are stored in a database. The micro-activities may include micro-activities and cluster identifiers (e.g., 314).

[0077] At 404, characteristic vectors are determined per micro-activity. The arithmetic average of selected features are used to create a single characteristic vector. The single vector includes average SOG, average delta SOG, and Average delta COG.

[0078] At 406, a classifier is trained using micro-activity characteristic vectors and cluster identifiers. The classifier may provide a regularizing gradient boosting framework. The classifier may provide any one or more of scalable, portable and distributed gradient boosting. The classifier may runs on a single machine and / or a distributed processing frameworks. The input for the classifier may include micro-activity characteristic vectors, each associated with a label. The labels may be the result of the clustering phase (e.g., noted at 310) using, for example, HDBSCAN.

[0079] The classifier may be an XGBoost (extreme Gradient Boosting) classifier. The “GBCIassifier” method may be from the “xgboost” PythonTM library.

[0080] At 408, classifier performance is validated using validation techniques. The validation techniques may include 10-fold cross validation and / or confusion matrix.

[0081] At 410, the trained classifier is stored.

[0082] Referring now to Figure 5, shown therein is flow chart of a method 500 for modeling and training, according to an embodiment. The method 500 may be performed the modeler and trainer 106 of Figure 1 .

[0083] At 502, a subset of end-to-end tracks with micro-activities is stored in a database. There may be four micro-activities including maneuvering, fishing, stationary, and steaming. The micro-activities may also include dark movement.

[0084] In contrast, TripTracker considered six types of activities, including two types of fishing, two types of being stationary, maneuvering, and steaming. Detailed explanations of the differences and characteristics of these activities, along with adiscussion on the findings related to them, are available in TripTracker. For segmentation and visualization purposes, the method 500 combines the two fishing activities and the two stationary activities. For the behavioral analysis of fishing vessels, seven activities may be considered.

[0085] At 504, configuration parameters are set up for a model. The model may be a stochastic model used to model pseudo-random ly changing systems. The model has it assumed that future states depend only on the current state, not on the events that occurred before it. Generally, this assumption enables reasoning and computation with the model that would otherwise be intractable. It may be desirable for the model to exhibit the Markov property.

[0086] The model may be a Hidden Markov Model (HMM). The HHM may include observations that are dependent on a latent (or "hidden") Markov process (referred to as X). The model includes an observable process Y, whose outcomes depend on the outcomes of X in a known way. Since X cannot be observed directly, the goal is to learn about state of X by observing Y. By definition of being a Markov model, an HMM has an additional requirement that the outcome of Y at time t = to must be "influenced" exclusively by the outcome of X at t = to and that the outcomes of X and Y at t<tO must be conditionally independent of Y at t = to given X at time t = to. Estimation of the parameters in an HMM can be performed using maximum likelihood. For linear chain HMMs, the Baum-Welch algorithm can be used to estimate the parameters.

[0087] The configuration parameters may include hidden states. The hidden states may include six hidden states, one stare per distinct cluster. The configuration parameters may include the SOG and COG features as observations.

[0088] The model may include models other than the HMM and its variations, such as the Gaussian Mixture Model HMM (GMMHMM). The HMM may be well-suited for modeling sequential data without labels, allowing the system to learn the underlying patterns in the data. The model may include the GMMHMM.

[0089] Apart from the HMM family, the model may be a sequential model having labeled data to classify the stream of data into multiple classes.

[0090] Utilizing HMM may identify activity transition points within a trip. Subsequently, applying the classifier (e.g., XGBoost) may determine the activity associated with each part of the trip. The combination of the model (e.g., HMM) with the classifier (e.g., XGBoost) segments each trip.

[0091] The output, segmented trips with activity labels, may be utilized to train other sequential classifier models. Where the method does not have access to data with activity labels, the method may include the combination of the model (e.g., HMM) and the classifier (e.g., XGBoost) to process unlabeled AIS data.

[0092] At 506, the model is trained to determine state transitions. The most likely sequence of states corresponding to a given sequence of observations (i.e. , a given trip) is derived from the trained model (e.g., HMM) using one of various algorithms such as the Viterbi algorithm. The predict X method in the model (e.g., HMM) is used to find the most likely state sequence for a given observation sequence X.

[0093] At 508, the trained model is stored.

[0094] At 510, the model transformers are stored.

[0095] Referring now to Figure 6, shown therein is flow chart of a method 600 for activity segmentation, according to an embodiment. The method 600 may be performed by the activity segmenter 108 of Figure 1 .

[0096] At 602, the pre-processed tracks (e.g., pre-processed tracks 210 of Figure 2), are received.

[0097] At 604, a trained model (e.g., trained model 508 of Figure 5) is received. At 606, model transformers (e.g., model transformers 510 of Figure 5) are received.

[0098] At 608, the trained model is used to infer state transition sequence for the track.

[0099] At 610, an inferred state transition sequence is added to the track.

[0100] At 612, a characteristic vector per segments is determined by same state per track. The arithmetic average of selected features are used to create a singlecharacteristic vector. The single vector may include the average SOG, the average delta SOG, and the average delta COG.

[0101] At 614, a trained classifier (e.g., trained classifier 410 of Figure 4) is received.

[0102] At 616, the trained classifier is used to map characteristic vectors of segments into activities.

[0103] At 618, the segmented tracks are written. The segmented tracks and the inputs and outputs may be provided or stored in a file format that operates for the system (e.g., CSV, JSON, etc.). The segmented track may be an AIS track.

[0104] At 620, the segmented track is stored.

[0105] The method 600 is illustrated for a single track. The method 600 may be with single or multiple tracks. The output of the method 600 corresponds to the given input (e.g., one or multiple tracks).

[0106] Referring now to Figure 7, shown therein is flow chart of a method 700 for track post processing, according to an embodiment. The method 700 may be performed by the track post processor 110 of Figure 1 .

[0107] At 702, a segmented track (e.g., segmented track 620 of Figure 6), is received.

[0108] At 704, visualization configuration parameters are setup. The parameters may include whether the interpolated points are displayed, or some smoothing on the resulting track to make activity segments contiguous. The visualizations include a plot of each data point (latitude and longitude) on the map, using a distinct color to represent each activity.

[0109] At 706, desired configuration is applied to the track.

[0110] At 708 the resulting track is written. The resulting track may be written to a computer file.

[0111] At 710 the output track is stored. The output track may be an AIS track.

[0112] The method 700 is illustrated for a single track. The method 700 may be with single or multiple tracks. The output of the method 700 corresponds to the given input (e.g., one or multiple tracks).

[0113] Figure 8 illustrates an image 800 including an AIS track, in accordance with an embodiment. The image 800 includes AIS position reports 802 (shown as dots). The AIS position reports 802 are sampled of AIS signals at a rate of one position report per minute. The image 800 illustrates the potential activities the fishing ship performs along its end-to-end trip color-coded as per the segmented activities found.

[0114] The green line 808 represents the positions at which the vessel is likely to be fishing. The blue line 812 represents the positions at which the vessel is likely to be steaming. The red lines 810 represent the positions where the vessel is likely to be stationary. The cyan lines 806 represents the positions where the vessel is likely to be maneuvering. The black lines 804 represents the positions where there is dark activity.

[0115] The activity includes dark movement 804 where there is a lack of AIS reports. The activity includes maneuvering 806 where the vessel is moving cautiously (e.g. leaving or entering a port). The activity includes fishing 808 where the vessel is engaging in fishing activity. The activity includes stationary 810 where the vessel is idle or berthed (e.g., engines are turned off). The activity includes steaming 812 where the vessel is traveling at higher speeds.

[0116] 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 system for mapping vessel track behaviour, the system comprising: an activity detector for detecting vessel activity of a track; a modeler and trainer for modeling and classifying vessel activity of the track, wherein the modeler and trainer is configured to: set up configuration parameters for a model; and train the model to determine state transitions; and a track activity segementer for segmenting vessel activity of the track, wherein the trained model is used to infer a state transition sequence for the track, wherein the track activity segementer is configured to: add the inferred state transition sequence to the track; determine characteristic vector per segments by same state per track; and map characteristic vectors of segments into activities with a trained classifier.

2. The system of claim 1 , wherein the model is a hidden Markov model (HMM).

3. The system of claim 1 , where the configuration parameters include any one or more of hidden states, speed over ground (SOG) and course over ground (COG) features as observations.

4. The system of claim 1 , wherein the activity detector is configured to: determine a timer interval for micro-activities;shred pre-processed tracks into micro-activity tables; and determine a characteristic vector per micro-activity.

5. The system of claim 1 , wherein the activity detector is configured to: cluster a subset of characteristic vectors to identify distinct activities; and validate the cluster analysis by running a clustering algorithm on different sample data.

6. The system of claim 1 , wherein the activity detector is configured to: determine characteristic vectors per micro-activity; train a classifier using micro-activity characteristic vectors and cluster identifiers; and validate classifier performance using validation techniques.

7. The system of claim 1 further comprising: a track preprocessor for pre-processing the track, wherein the track preprocessor is configured to: validate the format for the track and the features of the track; and interpolate the track features to one position per a given time period.

8. The system of claim 1 wherein the track is an AIS track, and the AIS track includes any one or more of latitude, longitude, time stamp, speed over ground (SOG), course over ground (COG), and maritime mobile service identity (MMSI).

9. The system of claim 1 further comprising: track post processor for post processing the track, wherein the track post processor is configured to: set up visualization configuration parameters; and apply desired configuration to the track.

10. The system of claim 1 , wherein the activity includes any one or more of maneuvering, fishing, stationary, and steaming.

11. A method for mapping vessel track behaviour, the method comprising: detecting a vessel activity of a track; modeling and classifying vessel activity of the track including: setting up configuration parameters for a model; and training the model to determine state transitions; and segmenting vessel activity of the track, wherein the trained model is used to infer a state transition sequence for the track, wherein segmenting vessel activity of the track includes: adding the inferred state transition sequence to the track; determining characteristic vector per segments by same state per track; and mapping characteristic vectors of segments into activities.

12. The method of claim 11 , wherein the model is a hidden Markov model (HMM).

13. The method of claim 11 , where the configuration parameters include any one or more of hidden states, speed over ground (SOG) and course over ground (COG) features as observations.

14. The method of claim 11 further comprising: determining a timer interval for micro-activities; shredding pre-processed tracks into micro-activity tables; and determining a characteristic vector per micro-activity.

15. The method of claim 11 further comprising: clustering a subset of characteristic vectors to identify distinct activities; and validating the cluster analysis by running a clustering algorithm on different sample data.

16. The method of claim 11 further comprising: determining characteristic vectors per micro-activity; training a classifier using micro-activity characteristic vectors and cluster identifiers; and validating classifier performance using validation techniques.

17. The method of claim 11 further comprising: validating the format for the track and the features of the track; and interpolating the track features to one position per a given time period.

18. The method of claim 11 wherein the track is an AIS track, and the AIS track includes any one or more of latitude, longitude, time stamp, speed over ground (SOG), course over ground (COG), and maritime mobile service identity (MMSI).

19. The method of claim 11 further comprising: setting up visualization configuration parameters; and applying desired configuration to the track.

20. The method of claim 11 , wherein the activity includes any one or more of maneuvering, fishing, stationary, and steaming.

Images