System And Method For Motion Abstraction, Activity Identification, And Vehicle Classification

• US 20200133961A1
• Filed: 10/31/2018
• Published: 04/30/2020
• Est. Priority Date: 10/31/2018
• Status: Active Grant

• Current Assignee: the united states of america as represented by the secretary of the navy
• Original Assignee: the united states of america as represented by the secretary of the navy
• Inventors: Wroblewski, Ronald J., Walton, Michael W.
• Granted Patent: US 11,151,169 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 16/287 Visualization; Browsing
• CPC Class Codes: G06N 20/20 Ensemble learning
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent
• CPC Class Codes: G06N 5/01 Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
• CPC Class Codes: G06N 5/022 Knowledge engineering; Knowledge acquisition
• CPC Class Codes: G08G 3/00 Traffic control systems for marine craft marking of navigational route B63B22/16, B63B51/00

Abstract

Motion abstraction includes ontologies having taxonomies for classification of various types of vessels (entities) and their movements based on inputted raw data. Kinematic-data abstraction, activity identification, entity classification, and entity identification, can be performed such that kinematic data is decomposed using an ontology describing motion, activities are decomposed using an ontology describing activities, and entity classes are decomposed using an ontology describing entity classes having unique-entity instances.

Claims

1. A system for kinematic-data abstraction, activity identification, entity classification, and entity identification, wherein kinematic data is decomposed using an ontology describing motion, activities are decomposed using an ontology describing activities, and entity classes are decomposed using an ontology describing entity classes having unique-entity instances, the system comprising:

a communication component configured to receive the kinematic data, which includes sets of data each including an entity tag, a timestamp, and a position vector;

a controller configured to sequentially process the kinematic data utilizing the ontology describing motion, which includes allowed motion states for a domain of interest, and wherein the sequential processing provides sequences of motion states with metadata including motion details;

a determining component configured to process the sequences of motion states, wherein the processing includes;

identifying an activity of an entity utilizing the ontology describing activities,

classifying an entity class utilizing the ontology describing entity classes, and

identifying the entity utilizing archived motion-state sequences; and

a display configured to output results showing motion-abstract patterns with relative distributions of the patterns ordered within an ontology and taxonomy based on the decomposed kinematic data.

2. The system of claim 1, wherein the motion details include at least four of:

time, position, velocity, proximity to other entities, and proximity to other regions.

3. The system of claim 1, wherein the kinematic data is provided in a form chosen from at least one of:

track data, AIS data, ASD-B data, and geolocation and time (GLT) data; and

wherein the sets of data further include a velocity vector.

4. The system of claim 1, wherein the kinematic-data abstraction is based at least in part on utilizing tolerances and thresholds of motion states to determine between whether a datum in one of the sets of data is consistent with a current motion state and whether the datum is indicative of a new motion state.

5. The system of claim 1, wherein the kinematic-data abstraction is based at least in part on utilizing dead-reckoning techniques from time-stamped position data, and wherein velocity data is not utilized.

6. The system of claim 1, wherein the kinematic-data abstraction is based at least in part on utilizing a taxonomy of the ontology describing motion, wherein the taxonomy is constrained by specific ontological exclusions when a particular motion state is followed by another particular motion state.

7. The system of claim 6, wherein the ontology describing motion pertains to the maritime domain, and wherein the motion states are constrained to a marine surface and physically allowed positions, velocities, and accelerations.

8. A method for kinematic-data abstraction, activity identification, entity classification, and entity identification, wherein kinematic data is decomposed using an ontology describing motion, activities are decomposed using an ontology describing activities, and entity classes are decomposed using an ontology describing entity classes having unique-entity instances, the method comprising:

receiving the kinematic data via a communication component, wherein the kinematic data comprises sets of data each including an entity tag, a timestamp, and a position vector;

sequentially processing the kinematic data via a controller, wherein the sequential processing utilizes the ontology describing motion, which includes allowed motion states for a domain of interest, and wherein the sequential processing provides sequences of motion states with metadata including motion details;

processing the sequences of motion states, via a determining component, wherein the processing includes;

identifying an activity of an entity utilizing the ontology describing activities,

classifying an entity class utilizing the ontology describing entity classes, and

identifying the entity utilizing archived motion-state sequences; and

outputting results to a display showing motion-abstract patterns with relative distributions of the patterns ordered within an ontology and taxonomy based on the decomposed kinematic data.

9. The method of claim 8, wherein the motion details include at least four of:

time, position, velocity, proximity to other entities, and proximity to other regions.

10. The method of claim 8, wherein the entity tag identifies a sequence of kinematic data for a single entity.

11. The method of claim 8, wherein the kinematic data is provided in a form chosen from at least one of:

track data, AIS data, ADS-B data, and geolocation and time (GLT) data; and

wherein the sets of data further include a velocity vector.

12. The method of claim 8, wherein the kinematic data is provided as velocity data that is entity-tagged and time-stamped, and wherein relative positions the entity are derived from the velocity data.

13. The method of claim 8, wherein the kinematic-data abstraction is based at least in part on utilizing tolerances and thresholds of motion states to determine between whether a datum in one of the sets of data is consistent with a current motion state and whether the datum is indicative of a new motion state.

14. The method of claim 8, wherein the kinematic-data abstraction is based at least in part on utilizing dead-reckoning techniques from time-stamped position data, and wherein velocity data is not utilized.

15. The method of claim 8, wherein the kinematic-data abstraction is based at least in part on utilizing a taxonomy of the ontology describing motion, wherein the taxonomy is constrained by specific ontological exclusions when a particular motion state is followed by another particular motion state.

16. The method of claim 15, wherein the ontology describing motion pertains to the maritime domain, and wherein the motion states are constrained to a marine surface and physically allowed positions, velocities, and accelerations.

17. The method of claim 15, wherein the ontology describing motion pertains to the ground domain, and wherein the motion states are constrained to a local topology and physically allowed positions, velocities, and accelerations.

18. The method of claim 15, wherein the ontology describing motion pertains to the aerial domain, and wherein the motion states are constrained a local topology, atmospheric height, and physically allowed positions, velocities, and accelerations.

19. The method of claim 8, wherein the processing the sequences of motion states is obtained via a machine-learning technique chosen from at least one of:

likelihood methods, decision trees, random forests, neural networks, and recurrent neural networks (RNNs).

20. The method of claim 8, wherein the processing the sequences of motion states is obtained based at least in part on one of:

activity models and entity-class models.