CN113658214B - Trajectory prediction method, collision detection method, apparatus, electronic device, and medium - Google Patents
Trajectory prediction method, collision detection method, apparatus, electronic device, and medium Download PDFInfo
- Publication number
- CN113658214B CN113658214B CN202110940464.0A CN202110940464A CN113658214B CN 113658214 B CN113658214 B CN 113658214B CN 202110940464 A CN202110940464 A CN 202110940464A CN 113658214 B CN113658214 B CN 113658214B
- Authority
- CN
- China
- Prior art keywords
- target object
- target
- predicted
- track
- trajectory
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 68
- 238000001514 detection method Methods 0.000 title claims abstract description 25
- 230000003068 static effect Effects 0.000 claims abstract description 32
- 239000013598 vector Substances 0.000 claims description 22
- 238000012545 processing Methods 0.000 claims description 20
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000013527 convolutional neural network Methods 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 7
- 238000007499 fusion processing Methods 0.000 claims description 7
- 230000008569 process Effects 0.000 claims description 6
- 230000004927 fusion Effects 0.000 claims description 5
- 238000012935 Averaging Methods 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000013473 artificial intelligence Methods 0.000 abstract description 3
- 238000004891 communication Methods 0.000 description 18
- 238000010586 diagram Methods 0.000 description 15
- 238000004590 computer program Methods 0.000 description 11
- 238000004422 calculation algorithm Methods 0.000 description 7
- 230000007246 mechanism Effects 0.000 description 7
- 230000000694 effects Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 230000010354 integration Effects 0.000 description 4
- 230000007774 longterm Effects 0.000 description 4
- 238000007726 management method Methods 0.000 description 4
- 238000012544 monitoring process Methods 0.000 description 4
- 238000001914 filtration Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 230000008447 perception Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000013500 data storage Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 235000019580 granularity Nutrition 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012806 monitoring device Methods 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000001953 sensory effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 230000003313 weakening effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/207—Analysis of motion for motion estimation over a hierarchy of resolutions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Traffic Control Systems (AREA)
Abstract
The present disclosure discloses a trajectory prediction method, a collision detection apparatus, an electronic device, a storage medium, and a program product, and relates to the technical field of artificial intelligence, and in particular, to the technical field of spatiotemporal big data. The specific implementation scheme is as follows: determining a first predicted trajectory of the target object based on the sequence of dynamic features and the static features associated with the target object; determining a second predicted trajectory of the target object based on the historical trajectory of the target object; and determining a target predicted trajectory of the target object based on the first predicted trajectory and the second predicted trajectory.
Description
Technical Field
The present disclosure relates to the field of artificial intelligence technologies, and in particular, to the field of spatiotemporal big data technologies, and in particular, to a trajectory prediction method, a collision detection method, an apparatus, an electronic device, a storage medium, and a program product.
Background
With the continuous application of modern science and technology to the shipping field, the upsizing, the novelty and the automation of ships are continuously improved. Higher requirements and standards are also provided for the aspects of shipping control, maintenance management and the like of the ship so as to improve the identification and control capability of the shipping risks.
Disclosure of Invention
The present disclosure provides a trajectory prediction method, a collision detection method, an apparatus, an electronic device, a storage medium, and a program product.
According to an aspect of the present disclosure, there is provided a trajectory prediction method including: determining a first predicted trajectory of the target object based on the sequence of dynamic features and the static features associated with the target object; determining a second predicted trajectory of the target object based on the historical trajectory of the target object; and determining a target predicted trajectory of the target object based on the first predicted trajectory and the second predicted trajectory.
According to another aspect of the present disclosure, there is provided a collision detection method including: determining respective target predicted trajectories of a plurality of objects within a target region; determining whether at least two intersected target prediction tracks exist in the target prediction tracks of the objects; and determining that there is a risk of collision for at least two objects corresponding to the at least two intersecting target predicted trajectories if it is determined that there are at least two intersecting target predicted trajectories; the target prediction track is obtained by prediction by using a track prediction method.
According to another aspect of the present disclosure, there is provided a trajectory prediction apparatus including: a first determination module to determine a first predicted trajectory of the target object based on the sequence of dynamic features and the static features associated with the target object; a second determination module for determining a second predicted trajectory of the target object based on the historical trajectory of the target object; and a third determination module for determining a target predicted trajectory of the target object based on the first predicted trajectory and the second predicted trajectory.
According to another aspect of the present disclosure, there is provided a collision detection apparatus including: a fourth determining module, configured to determine respective target predicted trajectories of multiple objects within the target region; a fifth determining module, configured to determine whether at least two intersected target predicted trajectories exist in respective target predicted trajectories of the multiple objects; and a sixth determining module for determining that there is a risk of collision of at least two objects corresponding to the at least two intersecting target predicted trajectories, if it is determined that there are at least two intersecting target predicted trajectories; the target predicted trajectory is predicted by using the trajectory prediction method described above.
According to another aspect of the present disclosure, there is provided an electronic device including: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method as described above.
According to another aspect of the present disclosure, there is provided a non-transitory computer readable storage medium having stored thereon computer instructions for causing the computer to perform the method as described above.
According to another aspect of the present disclosure, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the method as described above.
It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present disclosure, nor do they limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description.
Drawings
The drawings are included to provide a better understanding of the present solution and are not to be construed as limiting the present disclosure. Wherein:
FIG. 1 schematically illustrates an exemplary system architecture to which the trajectory prediction method and apparatus may be applied, according to an embodiment of the present disclosure;
FIG. 2 schematically illustrates a flow diagram of a trajectory prediction method according to an embodiment of the present disclosure;
FIG. 3 schematically shows an architecture diagram of a codec model according to an embodiment of the disclosure;
FIG. 4 schematically illustrates an architecture diagram for multi-model integration in accordance with an embodiment of the disclosure;
FIG. 5 schematically shows a trellis diagram for one-dimensional encoding according to an embodiment of the disclosure;
FIG. 6 schematically illustrates displaying a history track on a one-dimensional encoded mesh according to an embodiment of the disclosure;
FIG. 7 schematically shows a grid schematic of a two-dimensional code according to an embodiment of the disclosure;
FIG. 8 schematically illustrates a flow chart of a collision detection method according to an embodiment of the present disclosure;
FIG. 9 schematically illustrates a flow chart of a collision detection method according to another embodiment of the present disclosure;
FIG. 10 schematically illustrates a block diagram of a trajectory prediction device according to an embodiment of the present disclosure;
fig. 11 schematically shows a block diagram of a collision detection apparatus according to an embodiment of the present disclosure; and
fig. 12 schematically shows a block diagram of an electronic device adapted to implement a trajectory prediction method or a collision detection method according to an embodiment of the present disclosure.
Detailed Description
Exemplary embodiments of the present disclosure are described below with reference to the accompanying drawings, in which various details of the embodiments of the disclosure are included to assist understanding, and which are to be considered as merely exemplary. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.
In recent years, with the continuous development of fishery informatization, the upgrading and transformation investment of ship navigation equipment is continuously increased, the popularization and application of a satellite communication navigation technology in ship monitoring and dispatching are increased, the management means of fishery production safety is effectively improved, and the marine ship accidents generally fall. However, fishery safety is still the outstanding short plate in current fishery development, ship safety supervision is still the difficulty of fishery work, and ship marine accidents still occur occasionally, so that the safety of fishermen's lives and properties is seriously threatened. The reason is manifold, but still there are many weak links in the aspect of boats and ships safety information management and control, mainly show in:
the ship early warning capability is insufficient in advance, and the fishery safety management intellectualization degree is not high. The operation and safety guarantee of the ship is weak, the risk perception capability is insufficient, the early warning capability is low, and particularly, island reefs around the ship, other ships, unpredictable meteorological sea conditions and the like cannot be found in time at night.
The embodiment of the disclosure provides a track prediction method, a collision detection method, a device, electronic equipment, a storage medium and a program product.
According to an embodiment of the present disclosure, a trajectory prediction method may include: determining a first predicted trajectory of the target object based on the sequence of dynamic features and the static features associated with the target object; determining a second predicted trajectory of the target object based on the historical trajectory of the target object; and determining a target predicted trajectory of the target object based on the first predicted trajectory and the second predicted trajectory.
According to an embodiment of the present disclosure, a collision detection method may include: determining respective target predicted trajectories of a plurality of objects within a target region; determining whether at least two intersected target prediction tracks exist in the target prediction tracks of the objects; and determining that there is a risk of collision for at least two objects corresponding to the at least two intersecting target predicted trajectories if it is determined that there are at least two intersecting target predicted trajectories; the target prediction track is obtained by prediction by using a track prediction method.
By using the track prediction method provided by the embodiment of the disclosure, the final target predicted track can be determined by simultaneously combining the first predicted track and the second predicted track, so that the accuracy of the track prediction result is improved, and the application range is wider, namely, the navigation track of the short-term route is accurately predicted, and the method is also applicable to the long-term route. Therefore, the target prediction track predicted by the track prediction method provided by the embodiment of the disclosure is used for detecting ship collision, the early warning capability and the risk perception capability are improved, and the safety of marine navigation is improved.
In the technical scheme of the disclosure, the acquisition, storage, application and the like of the personal information of the related user all accord with the regulations of related laws and regulations, and do not violate the good customs of the public order.
FIG. 1 schematically illustrates an exemplary system architecture to which the trajectory prediction method and apparatus may be applied, according to an embodiment of the present disclosure.
It should be noted that fig. 1 is only an example of a system architecture to which the embodiments of the present disclosure may be applied to help those skilled in the art understand the technical content of the present disclosure, and does not mean that the embodiments of the present disclosure may not be applied to other devices, systems, environments or scenarios.
As shown in fig. 1, the system architecture 100 according to this embodiment may include a vessel 110, satellite communications 120, a server 130. Satellite communications 120 is the medium used to provide a communication link between the vessel 110 and the server 130. Satellite communications 120 may be communications between radio communication stations using a satellite as a relay.
The server 130 may interact with the vessel 110 via satellite communications 120 to receive or transmit positioning navigation messages, alerts, and the like.
The vessel 110 may be a navigation device such as a fishing boat, a commercial boat, a speed boat, etc.
According to the embodiment of the present disclosure, the ship 110 may have a Positioning navigation device installed thereon, including but not limited to a GPS (Global Positioning System), a compass, an AIS (Automatic Identification System), and the like. The vessel 110 may further include auxiliary monitoring devices such as a speed monitor and a pilot, which are related to navigation.
The server 130 may be a server that provides various services. The server 130 may include a data storage module, a data processing module, and a data query module. The data storage module can store data related to navigation sent by the ship, wherein the data related to navigation can be related to the current navigation of the ship or related to a historical course of the ship. The data query module can be a query module for data of weather, sea area environment and the like, and can also be a query module for attribute characteristics of ships. The data processing module can be used for predicting the target prediction track of the ship based on the data related to navigation sent by the ship and the data obtained by the query of the data query module.
It should be noted that the trajectory prediction method provided by the embodiment of the present disclosure may be generally executed by a processor installed on the ship 110. Accordingly, the trajectory prediction device provided by the embodiment of the present disclosure may also be disposed in the ship 110.
Alternatively, the trajectory prediction method provided by the embodiment of the present disclosure may also be generally executed by the server 130. Accordingly, the trajectory prediction device provided by the embodiment of the present disclosure may be generally disposed in the server 130.
It should be understood that the number of vessels, satellite communications, and servers in fig. 1 are merely illustrative. There may be any number of vessels, satellite communications, and servers, as desired for implementation.
FIG. 2 schematically shows a flow chart of a trajectory prediction method according to an embodiment of the present disclosure.
As shown in fig. 2, the method includes operations S210 to S230.
In operation S210, a first predicted trajectory of the target object is determined based on the sequence of dynamic features and the static features associated with the target object.
In operation S220, a second predicted trajectory of the target object is determined based on the historical trajectory of the target object.
In operation S230, a target predicted trajectory of the target object is determined based on the first predicted trajectory and the second predicted trajectory.
According to the embodiment of the present disclosure, the type of the target object is not limited, and may be a merchant ship, a fishing boat, or a yacht, for example.
According to the embodiments of the present disclosure, a marine navigation environment, which has no developed road similar to land and no night auxiliary lighting, and weather sea conditions are variable and unpredictable. Therefore, the marine vessel has low risk perception capability and early warning capability to the surrounding driving environment. Therefore, it is necessary to predict the trajectory of the object traveling on the sea to improve the understanding of the surrounding environment.
According to an embodiment of the present disclosure, the dynamic feature sequence associated with the target object may be a dynamic feature sequence of a single factor or a dynamic feature sequence of multiple factors. For example, the characteristic sequence may be a single characteristic sequence of the navigation track point, or may be a dynamic characteristic sequence of a set of driving states related to navigation, such as speed, heading, track point, and the like, and may further include other external influence factors in addition to the dynamic characteristic sequence of the driving states related to navigation, such as a dynamic characteristic sequence of environment, weather, and the like. Any dynamic feature sequence that affects navigation with respect to the target object may be used.
According to an embodiment of the present disclosure, the dynamic feature sequence associated with the target object may include a historical dynamic feature sequence for the current voyage, for example, a dynamic feature sequence associated with the target object that has occurred in the current time period.
According to an embodiment of the present disclosure, the static feature may be a feature that does not change over time, such as an attribute feature of the target object, but is not limited thereto, and may also be other features related to the target object.
According to an embodiment of the present disclosure, the first predicted trajectory may be a trajectory predicted by using the current navigation, the dynamic feature sequence of the current navigation, and the static feature. The first predicted track obtained by the data characteristic prediction of the current navigation is high in execution efficiency and good in real-time performance, and is suitable for prediction of short-term route tracks.
According to embodiments of the present disclosure, the historical trajectory may be a historical course trajectory over a period of time of the target object, for example, a historical course trajectory within a month of the current month. The history track may be one or more, and is not limited.
According to an embodiment of the present disclosure, the second predicted trajectory may be a trajectory predicted using a historical trajectory. And the second predicted track obtained by the historical track prediction is suitable for predicting the long-term track of the ship driven by the general air route.
According to the embodiment of the disclosure, the target predicted track can be determined based on the first predicted track and the second predicted track, and not only the real-time performance of the short-term route but also the universality and fixedness of the long-term route are considered. The short-term route and the long-term route are combined, so that the application range is wider, and the accuracy of the trajectory prediction is high.
A method such as that shown in fig. 2 is further described below with reference to fig. 3-7 in conjunction with specific embodiments.
According to the embodiment of the disclosure, the prediction of the short-term track of the ship as the target object can be regarded as a position fitting problem, namely, the position of the ship in the future is predicted based on the existing track, the driving state and the environmental factors in the current time period.
According to the embodiment of the disclosure, considering that the track is a position sequence formed in time sequence, the adjacent position points have a larger relevance. Accordingly, embodiments of the present disclosure employ a dynamic feature, i.e., a sequence of dynamic features, associated with a target object that has timing information.
According to embodiments of the present disclosure, the dynamic feature sequence may include speed, heading, location, weather, time, surrounding environment, and the like.
According to an embodiment of the present disclosure, the dynamic signature sequence may include a driving state of each time point that has been generated within a current time period (e.g., within the past T hours), e.g., { T } 1 、T 2 、T 3 、…T k Speed, heading, and location of time.
According to embodiments of the present disclosure, the weather may be, for example, sunny, rainy, windy, ocean current, waves, and the like. The weather of the dynamic feature sequence in the embodiment of the present disclosure may be a weather feature sequence recorded in time series over time.
According to an embodiment of the present disclosure, the surrounding environment may refer to an environment around a location where the target object is located. Such as whether there is a ship or other obstruction within a safe distance, etc. The ambient environment of the dynamic feature sequence in the embodiment of the present disclosure may be an environmental feature sequence recorded in time series over time.
According to the embodiment of the present disclosure, time may be a time dimension such as morning, afternoon, evening, and the like, and may also refer to a time dimension such as 1 point, 2 points, and the like.
According to the embodiment of the disclosure, the prediction of the first predicted track can not only embody timeliness by using the running state of the existing track of the target object in the current time period of the current navigation, but also play a reference role by using the historical track of the target object.
For example, the existing track of the target object is subjected to mesh coding to obtain a coding result. And processing the coding result to obtain a semantic vector corresponding to the existing track of the target object. Determining the implicit semantic features of the existing track of the target object based on the semantic vector corresponding to the existing track of the target object, and taking the implicit semantic features of the existing track of the target object as a factor of the dynamic feature sequence.
According to the embodiment of the disclosure, the physical space may be gridded and each grid may be numbered, resulting in a grid code. Then, traversing the existing track of the target object, determining a coding result corresponding to the existing track, and then processing the coding result by using, for example, a word2vec algorithm, thereby obtaining a semantic vector of the existing track and further obtaining an implicit semantic feature.
According to another embodiment of the present disclosure, the physical space may also be gridded and each grid is numbered, resulting in a grid code. And then modeling the grids by using a word2vec algorithm so as to obtain a semantic vector of each grid. And traversing the existing track of the target object, and constructing the implicit semantic features of the track based on the semantic vector of each grid in the track.
It should be noted that the existing trajectory of the target object may be a trajectory that has been generated by the target object in the current time period, or may be a historical trajectory of a historical route of the target object. As long as the existing track of the target object exists, the position feature sequence can be embodied by the implicit semantic features.
According to an embodiment of the present disclosure, a static feature related to the target object may also be considered as a reference factor of the first predicted trajectory.
According to embodiments of the present disclosure, the static characteristics may include attribute characteristics of the target object, such as a specification, size, model, role, etc. of the ship.
According to the embodiment of the disclosure, not only the dynamic characteristic sequence associated with the target object but also the static characteristic of the target object are considered, and all factors are integrated and considered comprehensively, so that the accuracy of the first predicted trajectory is improved.
According to an embodiment of the present disclosure, a first predicted trajectory of a target object may be determined using a codec model based on a sequence of dynamic features and static features associated with the target object.
Fig. 3 schematically shows an architecture diagram of a codec model according to an embodiment of the present disclosure.
As shown in fig. 3, the encoding and decoding model 300 provided by the embodiment of the present disclosure may include an encoder 310 and a decoder 320. The sequence of dynamic features and static features associated with the target object may be input into a convolution module of the encoder 310, resulting in abstract features. The abstract features are input into a feature extraction module of the encoder 310 to obtain semantic vectors. The semantic vector is then input into the decoder 320, resulting in a first predicted trajectory.
According to embodiments of the present disclosure, the convolution module may include one or more Convolutional Neural Networks (CNNs). Where the convolution module includes a plurality of convolutional neural networks, the plurality of convolutional neural networks may be in a parallel relationship. The convolution module is used for performing feature convolution calculation on the input dynamic feature sequence and the static feature to obtain the abstract feature.
According to embodiments of the present disclosure, the feature extraction module may include an encoded attention (attention) mechanism and an encoded gated round-robin unit (GRU). The coding attention mechanism can weight the characteristics of different time nodes according to actual conditions, such as highlighting the weight of important time nodes and weakening unimportant information. The coding gated-loop elements can be sequence modeled to capture important information in the time dimension.
It should be noted that, in other embodiments of the present disclosure, the feature extraction module is not limited to the combination of the encoded attention mechanism and the encoded gating cycle unit, and may also include the encoded attention mechanism and the encoded long-short term memory network.
According to the embodiment of the disclosure, the feature extraction module is designed in the encoder, so that more important information can be captured on the basis of adjusting the weight, and the obtained semantic vector has a good encoding effect.
According to embodiments of the present disclosure, a decoder may include a decode attention mechanism and a decode gating loop unit. In embodiments of the present disclosure, a sequence may be constructed in conjunction with a semantic vector using a decoding attention mechanism and a decoding gating loop unit.
According to an embodiment of the present disclosure, the decoder may further include an output module. In embodiments of the present disclosure, the output module may include a multi-layer sensor and an activation function. The output module performs nonlinear conversion on the sequence obtained by the decoding attention mechanism and the decoding gating circulation unit to form a final first prediction track (namely a track sequence).
According to another embodiment of the present disclosure, the decoder may include a multi-layer perceptron (MLP) or decoding convolution module for transforming the semantic vectors to form the final first predicted trajectory (i.e., trajectory points).
The coding and decoding model provided by the embodiment of the disclosure is used as a sequence model to utilize and process the dynamic characteristic sequence and the static characteristic, so that the first prediction track has good and accurate prediction effect.
According to an embodiment of the present disclosure, the first predicted trajectory of the target object may also be determined by means of multi-model integration (i.e., fusion) based on the dynamic feature sequence and the static features associated with the target object.
For example, a dynamic feature sequence and a static feature associated with a target object are respectively input into a plurality of base learners to obtain prediction results of the base learners; and inputting the prediction results of the plurality of base learners into the secondary learner to obtain a first prediction track of the target object.
According to an embodiment of the present disclosure, the sub learner and each of the plurality of base learners may respectively include: the coding and decoding model comprises one or more of a Gaussian process regression network, a grey network, a multilayer perceptron, a convolutional neural network, a long-short term memory network and a gated cyclic unit.
FIG. 4 schematically shows an architecture diagram for multi-model integration according to an embodiment of the disclosure.
As in the example shown in fig. 4, the base learner may include 3, the base learner 1410, the base learner 2420, and the base learner 3430. The network architecture of each base learner is a codec model, including an encoder and a decoder. The input of each base learner is the same, and the input is the dynamic feature sequence and the static feature related to the target object, so as to obtain the output feature of each of the 3 base learners, and the output feature of each of the 3 base learners is used as the input feature of the secondary learner 440, so as to obtain the first predicted trajectory of the output of the secondary learner 440.
The first prediction track is predicted by utilizing the multi-model integration mode provided by the embodiment of the disclosure, so that the characteristics of a plurality of networks can be integrated, and the prediction effect is better.
According to the embodiment of the disclosure, a plurality of historical tracks can be processed by using a track clustering algorithm, similar historical tracks are clustered together, and then prediction is performed based on the clustered historical tracks to obtain a second predicted track, namely a relatively universal 'course' is formed.
According to the embodiment of the present disclosure, track clustering is performed on a plurality of historical tracks, which may be performed as follows.
For example, grid coding is performed on a plurality of historical tracks to obtain respective coding results of the plurality of historical tracks; and clustering the plurality of historical tracks based on the respective coding results of the plurality of historical tracks, and determining at least one target historical track.
Fig. 5 schematically shows a mesh schematic of one-dimensional encoding according to an embodiment of the disclosure.
The historical navigation area of the target object can be gridded and each grid can be uniquely encoded. In the example shown in fig. 5, one-dimensional encoding may be performed, and the trellis encoding result is { x1, …, x9} in turn.
FIG. 6 schematically shows a diagram of displaying a history track on a one-dimensional encoded mesh according to an embodiment of the disclosure.
As shown in fig. 6, the encoding result of the historical track may be determined according to the encoding of the related mesh, for example, the mesh encoding result constructed by the historical track T1 is { x1, x4, x5, x8, x9, x6 }.
According to the embodiment of the disclosure, each historical track in the plurality of historical tracks can be traversed in sequence, and each historical track is encoded according to the grid through which the historical track passes, so that the encoding result of each historical track is obtained.
According to the embodiment of the disclosure, the plurality of historical tracks can be clustered based on the respective encoding results of the plurality of historical tracks, and at least one target historical track is determined. In the embodiment of the present disclosure, for the encoding result of the one-dimensional encoding, clustering may be performed by using a set similarity manner, for example, using a jaccard distance to measure the similarity degree of different historical tracks, and then clustering the historical tracks with the similarity degree greater than or equal to a similarity threshold.
According to the embodiment of the present disclosure, the area is gridded, and the gridding is not limited to one-dimensional encoding, and two-dimensional encoding, three-dimensional encoding, or the like may be performed.
Fig. 7 schematically shows a mesh diagram of two-dimensional encoding according to an embodiment of the present disclosure.
As shown in fig. 7, the region may be gridded and two-dimensionally encoded. The two-dimensional encoding result may be { (x-1, y-1), (x-1, y), (x-1, y +1), (x, y-1), (x, y), (x, y +1), (x +1, y-1), (x +1, y), (x +1, y +1) }.
According to the embodiment of the disclosure, the plurality of historical tracks can be clustered based on the respective encoding results of the plurality of historical tracks, and at least one target historical track is determined. In the embodiment of the present disclosure, for the encoding result of the two-dimensional encoding, clustering may be performed using the euclidean distance and/or the edit distance.
According to the embodiment of the disclosure, the calculation mode of the euclidean distance may be to sequentially calculate the distances between the coding point pairs corresponding to the 2 historical tracks; the sum of the distances between corresponding pairs of encoded points is then taken as the similarity between 2 tracks.
According to an embodiment of the present disclosure, one implementation of the edit distance EDR may be as shown in equation (1) below.
Where EDR (a, B) is the edit distance of track a and track B, and epsilon is the distance threshold (one implementation for calculating subboost, subboost can be as shown in equation (2) below), the physical meaning of EDR distance of track a and track B is the minimum number of operations (optional operations: insert, delete, and replace) to change track a to track B.
Wherein m and n are the lengths of the track A and the track B; given trajectory A { a1, a2, …, an } and trajectory B { B1, B2, …, bm }, head (A) denotes a1, rest (A) denotes { a2, …, an }, head (B) denotes B1, rest (B) denotes { B2, …, bm }.
According to the exemplary embodiment of the present disclosure, the set similarity determined by the encoding result of the one-dimensional encoding may be integrated with the euclidean distance and the edit distance determined by the encoding result of the two-dimensional encoding, and clustering may be performed by the integrated similarity.
For example, different weights are configured for each similarity result, and a comprehensive similarity is obtained. One implementation of the integrated similarity may be shown, for example, in equation (3) below.
S w1 set similarity + w2 euclidean distance + w3 edit distance; (3)
wherein w1, w2 and w3 are weights of set similarity, Euclidean distance and edit distance, respectively, and S is comprehensive similarity.
According to the embodiment of the disclosure, the multiple historical tracks are clustered by utilizing the comprehensive similarity, so that the clustering effect is more accurate.
According to the embodiment of the disclosure, the region is gridded, and the size of grid division, namely the length and the width, is related to the encoding result of the subsequent historical track. The smaller the grid division is, the closer the encoding result of the historical track is to the actual historical track, and the larger the calculation amount of the relative clustering is; the larger the meshing, the smaller the amount of computation of the clustering.
According to the exemplary embodiment of the disclosure, under the condition that the number of the historical tracks is huge, the areas can be subjected to grid division with different granularities.
For example, the mesh may be first divided into coarse-grained regions, and then the divided mesh may be subjected to one-dimensional encoding and two-dimensional encoding to obtain coarse-grained comprehensive similarity, and then the coarse-grained comprehensive similarity is clustered to obtain a coarse-grained hierarchical clustering result. Then, fine-grained division is carried out on the grids, one-dimensional coding and two-dimensional coding are carried out on the divided grids to obtain fine-grained comprehensive similarity, and clustering is carried out based on the fine-grained comprehensive similarity to obtain a fine-grained hierarchical clustering result.
According to the embodiment of the disclosure, the coarse-grained grids are used for carrying out primary clustering on the plurality of historical tracks, so that the subsequent calculation amount is simplified, and the fine-grained grids are used for clustering the results after primary clustering again, so that the clustering effect is improved.
According to the embodiment of the disclosure, the similarity between at least one target historical track and the existing track of the target object in the current time period can be further determined; and determining the target historical track with the similarity greater than or equal to a preset similarity threshold as a second predicted track of the target object.
According to the embodiment of the disclosure, the historical track can be abstracted into a plurality of valuable routes, namely the target historical track, through the clustering. When the second predicted track is predicted subsequently, the existing track of the target object in the current time period may be compared with the target historical track, the target historical track with the similarity higher than the preset similarity threshold is found, and the target historical track is used as the subsequent forward route of the target object.
According to the embodiment of the disclosure, the preset similarity threshold is not specifically limited, and can be drawn up according to actual conditions, and the larger the preset similarity threshold is, the closer the existing track in the current time period obtained according to the preset similarity threshold is to the target historical track.
According to the embodiment of the disclosure, the similarity calculation can be performed by using a naive Bayes algorithm, so that the relation between the target historical track and the existing track of the target object in the current time period can be better described.
For example, the existing track of the target object in the current time period is S, and Ri represents the ith entry mark history track, and the specific steps are as follows.
And obtaining the probability P (Ri | S) (i ═ 1.. n) that the existing track S belongs to different target historical tracks in the current time period based on a naive Bayes algorithm.
And acquiring the maximum belonging probability P (Ri _ max | S), and if the value is greater than a preset similarity threshold value P, determining that the rear part of the existing track S in the current time period advances along the route Ri _ max.
According to an embodiment of the present disclosure, the position of the current ship T minutes later may be predicted based on the course Ri _ max.
For example, a point p1 closest to the current target object and a next track point p2 of the point are found on the route Ri _ max. And taking the included angle between the track points p1 and p2 as the subsequent heading of a target object such as a ship. And calculating the position of the ship after 20 minutes of running according to the current navigational speed and the current course, namely a second predicted track.
It is also possible to find the flight segment part closest to the current vessel, for example, on the course Ri _ max. And taking the direction of the sailing segment part as the subsequent course of the ship. And calculating the position of the ship after 20 minutes of running according to the current navigational speed and the current course, namely a second predicted track.
According to an embodiment of the present disclosure, a target predicted trajectory of a target object may be determined based on a first predicted trajectory and a second predicted trajectory by the following operations.
For example, the first predicted trajectory and the second predicted trajectory are subjected to fusion processing to determine a target predicted trajectory of the target object.
According to an embodiment of the present disclosure, the fusion process may include one or more of a voting method, an averaging method, and a weighting method.
The prediction results of different models are fused by using the integrated learning strategy provided by the embodiment of the disclosure, so that fusion of the prediction results of a plurality of models is obtained, correction is performed again from a data level, and the accuracy of trajectory prediction is improved.
Fig. 8 schematically shows a flow chart of a collision detection method according to an embodiment of the present disclosure.
As shown in fig. 8, the method includes operations S810 through S830.
In operation S810, a target predicted trajectory of each of a plurality of objects within a target region is determined.
In operation S820, it is determined whether at least two intersected target predicted trajectories exist among the target predicted trajectories of the respective objects.
In operation S830, in a case where it is determined that there are at least two intersecting target predicted trajectories, it is determined that there is a risk of collision of at least two objects corresponding to the at least two intersecting target predicted trajectories; the target prediction track is obtained by prediction by using a track prediction method.
According to the embodiment of the disclosure, the position and the number of the target areas are not limited, and for example, the target areas can be one area where a fishing boat is frequently present or a plurality of areas where a boat collision accident frequently occurs. It is not described herein, but may be in a region within the monitoring range.
According to an embodiment of the present disclosure, an object refers to an object that travels within a target area or enters within the target area, and there is a collision with the object within the target area.
According to embodiments of the present disclosure, an object may refer to a merchant ship, fishing boat, speed boat, or other device that travels, floats on a target area. Any equipment which can be monitored and affects the safe operation can be taken as the object.
According to the embodiment of the disclosure, the target predicted track of each of the plurality of objects can be predicted by adopting the track prediction method provided by the embodiment of the disclosure based on the received real-time bit reporting data and the historical storage data of each object.
According to the embodiment of the disclosure, the target prediction tracks of the multiple objects are obtained through prediction, and each target prediction track represents the future navigation track of each object. In the event that it is determined that there are at least two intersecting target predicted trajectories, it may be determined that there is a risk of collision for at least two objects corresponding to the at least two intersecting target predicted trajectories. In the event that it is determined that there are not at least two intersecting target predicted trajectories, it may be determined that there is no risk of an object colliding.
According to the embodiment of the disclosure, the collision risk is determined by using the trajectory prediction method provided by the embodiment of the disclosure, and on the basis of high target prediction trajectory prediction accuracy of each object, the collision detection has high prediction accuracy and strong safety early warning capability.
A method such as that shown in fig. 8 is further described below in conjunction with specific embodiments and with reference to fig. 9.
According to an embodiment of the present disclosure, in order to reduce the amount of calculation, the manner of determining the plurality of objects of the target area may be determined by the following operation.
For example, the filtering determination is made in a regular manner. Firstly, filtering out the fishing boat entering a target area, further filtering out the ships and unknown equipment which are close to the fishing boat in the target area, and then extracting the respective dynamic characteristic sequence and static characteristic of the ships and equipment to carry out collision detection.
According to the embodiment of the disclosure, the position reporting information can be stored at the same time, so that the dynamic characteristics of the ship are formed.
According to the embodiment of the disclosure, the regional monitoring mainly filters the acquired ship position through regional boundary information to judge whether the ship enters a monitored target region, and carries out subsequent processing on the ship entering the monitored target region.
According to the embodiment of the disclosure, it is directly judged that the object is time-consuming not in the target area based on the position of the object. A ray can be drawn by using a ray projection algorithm by taking the position of the object as an original point, and whether the intersection point of the ray and the boundary of the target area is an odd number or an even number can be judged. If odd, it indicates that the object is within the target region, otherwise it indicates that the object is not within the target region.
According to the exemplary embodiment of the disclosure, the area judgment problem can be converted into a direct matching problem, so that the data processing timeliness is improved. For example, a target area to be monitored is subjected to mesh segmentation, and a mesh list of the area is maintained. When receiving the position reporting data, gridding the position of an object such as a ship, and judging whether the grid where the ship is located is in the regional network, so as to monitor and judge the target region.
According to the embodiment of the disclosure, the target object is screened from the plurality of objects in the target area, the surrounding environment of the target object in the target area is screened, and the method is more targeted and improves the processing capacity and efficiency.
In consideration of the high risk of safe operation of the fishing vessel, the fishing vessel may be selected as a target object from a plurality of objects. And then judging and confirming the collision risk of other commercial ships at the position of the fishing ship within a certain position range threshold value within a certain preset time period.
According to the embodiment of the disclosure, a grid where the latest position data of the ship is located in a certain time period can be provided, for the fishing ship entering the monitoring target area, other commercial ships where the position of the fishing ship in the target area is located in a certain position range threshold value in a certain preset time period can be obtained, and a commercial ship-fishing ship pair list is screened out through grid matching. And then carrying out track prediction on the objects in the list by the merchant ship-fishing ship, and determining a target prediction track.
According to the embodiment of the disclosure, the time period is tentatively 10 minutes in consideration of the time of reporting the position of the Beidou or other navigation equipment. Meanwhile, the ship speed and the 20-minute alarm time are integrated, and the position range threshold value is temporarily set to be 50 KM. However, the present invention is not limited to this, and may be modified in accordance with the actual situation.
Fig. 9 schematically shows a flow chart of a collision detection method according to another embodiment of the present disclosure.
As shown in fig. 9, the method may obtain vessel grid information 920 for a target vessel based on a vessel location 910 of the target vessel, such as a fishing vessel. And matching the ship grid information 920 with the target area grid information 930, and stopping subsequent operation if the matching is unsuccessful. And in case of successful matching, determining that the target ship is in the target area. A plurality of ships entering the target area are monitored to obtain a ship grid information list 940. Screening other ships or equipment 950 in accordance with the position of the target ship (for example, the distance is not more than 50KM within 10 minutes) from the ship grid information list 940, respectively predicting target predicted tracks 960 of the target ship and the other ships or equipment 950 by using a track prediction method, and determining collision risk probability 970 based on the target predicted tracks 960.
By utilizing the determination mode of the object in the target area provided by the embodiment of the disclosure, the processing efficiency can be improved, and the safety early warning capability can be improved.
FIG. 10 schematically shows a block diagram of a trajectory prediction device according to an embodiment of the present disclosure.
As shown in fig. 10, the trajectory prediction apparatus 1000 may include a first determination module 1010, a second determination module 1020, and a third determination module 1030.
A first determination module 1010 configured to determine a first predicted trajectory of the target object based on the sequence of dynamic features and the static features associated with the target object.
A second determining module 1020 for determining a second predicted trajectory of the target object based on the historical trajectory of the target object.
A third determining module 1030, configured to determine a target predicted trajectory of the target object based on the first predicted trajectory and the second predicted trajectory.
According to an embodiment of the present disclosure, the first determining module may include a first input unit, a second input unit, and a decoding unit.
The first input unit is used for inputting the dynamic characteristic sequence and the static characteristic associated with the target object into a convolution module of the encoder to obtain the abstract characteristic.
And the second input unit is used for inputting the abstract features into a feature extraction module of the encoder to obtain the semantic vector.
And the decoding unit is used for inputting the semantic vector into a decoder to obtain a first prediction track.
According to an embodiment of the present disclosure, the first determination module may include a third input unit, and a fourth input unit.
And a third input unit, configured to input the dynamic feature sequence and the static feature associated with the target object into the plurality of base learners, respectively, to obtain prediction results of the plurality of base learners.
And the fourth input unit is used for inputting the prediction results of the plurality of base learners into the secondary learner to obtain a first predicted track of the target object.
According to an embodiment of the present disclosure, the secondary learner and each of the plurality of base learners respectively include at least one of:
the system comprises a coding and decoding model comprising a Gaussian process regression network, a coding and decoding model comprising a gray network, a coding and decoding model comprising a multilayer perceptron, a coding and decoding model comprising a convolutional neural network and a long-short term memory network, and a coding and decoding model comprising a gating cycle unit.
According to an embodiment of the present disclosure, the history track may include a plurality.
According to an embodiment of the present disclosure, the second determination module may include an encoding unit, a clustering unit, a calculation unit, and a determination unit.
And the coding unit is used for carrying out grid coding on the plurality of historical tracks to obtain the respective coding results of the plurality of historical tracks.
And the clustering unit is used for clustering the plurality of historical tracks based on the respective coding results of the plurality of historical tracks and determining at least one target historical track.
And the calculating unit is used for determining the similarity between at least one target historical track and the existing track of the target object in the current time period.
And the determining unit is used for determining the target historical track with the similarity greater than or equal to a preset similarity threshold as a second predicted track of the target object.
According to an embodiment of the present disclosure, the third determination module may include a fusion unit.
And the fusion unit is used for performing fusion processing on the first predicted track and the second predicted track and determining a target predicted track of the target object.
According to an embodiment of the present disclosure, the fusion process includes at least one of:
voting, averaging, and weighting.
According to an embodiment of the disclosure, the dynamic signature sequence comprises at least one of:
speed, heading, location, weather, time, ambient environment.
According to an embodiment of the present disclosure. The static features include attribute features of the target object.
According to an embodiment of the present disclosure, the dynamic feature sequence further comprises implicit semantic features.
According to an embodiment of the present disclosure, the trajectory prediction apparatus may further include an encoding module, a processing module, and a converting module.
And the coding module is used for carrying out grid coding on the existing track of the target object to obtain a coding result.
And the processing module is used for processing the coding result to obtain a semantic vector corresponding to the existing track of the target object.
And the conversion module is used for determining the implicit semantic features of the existing track of the target object based on the semantic vector.
Fig. 11 schematically shows a block diagram of a collision detection apparatus according to an embodiment of the present disclosure.
As shown in fig. 11, the collision detecting apparatus 1100 may include a fourth determining module 1110, a fifth determining module 1120, and a sixth determining module 1130.
A fourth determining module 1110, configured to determine target predicted trajectories of the multiple objects in the target area.
A fifth determining module 1120, configured to determine whether there are at least two intersecting target predicted trajectories among the target predicted trajectories of the respective plurality of objects.
A sixth determining module 1130, configured to determine that there is a risk of collision for at least two objects corresponding to at least two intersecting target predicted trajectories if it is determined that there are at least two intersecting target predicted trajectories; the target prediction track is obtained by prediction by the track prediction method.
The present disclosure also provides an electronic device, a readable storage medium, and a computer program product according to embodiments of the present disclosure.
According to an embodiment of the present disclosure, an electronic device includes: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method as described above.
According to an embodiment of the present disclosure, a non-transitory computer readable storage medium having stored thereon computer instructions for causing a computer to perform the method as described above.
According to an embodiment of the disclosure, a computer program product comprising a computer program which, when executed by a processor, implements the method as described above.
FIG. 12 shows a schematic block diagram of an example electronic device 1200, which can be used to implement embodiments of the present disclosure. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit implementations of the disclosure described and/or claimed herein.
As shown in fig. 12, the apparatus 1200 includes a computing unit 1201 which can perform various appropriate actions and processes in accordance with a computer program stored in a Read Only Memory (ROM)1202 or a computer program loaded from a storage unit 1208 into a Random Access Memory (RAM) 1203. In the RAM 1203, various programs and data required for the operation of the device 1200 may also be stored. The computing unit 1201, the ROM 1202, and the RAM 1203 are connected to each other by a bus 1204. An input/output (I/O) interface 1205 is also connected to bus 1204.
Various components in the device 1200 are connected to the I/O interface 1205 including: an input unit 1206 such as a keyboard, a mouse, or the like; an output unit 1207 such as various types of displays, speakers, and the like; a storage unit 1208, such as a magnetic disk, optical disk, or the like; and a communication unit 1209 such as a network card, modem, wireless communication transceiver, etc. The communication unit 1209 allows the device 1200 to exchange information/data with other devices via a computer network such as the internet and/or various telecommunication networks.
The computing unit 1201 may be a variety of general purpose and/or special purpose processing components having processing and computing capabilities. Some examples of the computing unit 1201 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various specialized Artificial Intelligence (AI) computing chips, various computing units running machine learning model algorithms, a Digital Signal Processor (DSP), and any suitable processor, controller, microcontroller, and so forth. The calculation unit 1201 executes the respective methods and processes described above, such as the trajectory prediction method or the collision detection method. For example, in some embodiments, the trajectory prediction method or the collision detection method may be implemented as a computer software program tangibly embodied in a machine-readable medium, such as storage unit 1208. In some embodiments, part or all of the computer program may be loaded and/or installed onto the device 1200 via the ROM 1202 and/or the communication unit 1209. When the computer program is loaded into the RAM 1203 and executed by the computing unit 1201, one or more steps of the trajectory prediction method or the collision detection method described above may be performed. Alternatively, in other embodiments, the computing unit 1201 may be configured to perform a trajectory prediction method or a collision detection method by any other suitable means (e.g., by means of firmware).
Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuitry, Field Programmable Gate Arrays (FPGAs), Application Specific Integrated Circuits (ASICs), Application Specific Standard Products (ASSPs), system on a chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, receiving data and instructions from, and transmitting data and instructions to, a storage system, at least one input device, and at least one output device.
Program code for implementing the methods of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the program codes, when executed by the processor or controller, cause the functions/operations specified in the flowchart and/or block diagram to be performed. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.
In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic, speech, or tactile input.
The systems and techniques described here can be implemented in a computing system that includes a back-end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), Wide Area Networks (WANs), and the Internet.
The computer system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server may be a cloud server, a server of a distributed system, or a server with a combined blockchain.
It should be understood that various forms of the flows shown above may be used, with steps reordered, added, or deleted. For example, the steps described in the present disclosure may be executed in parallel, sequentially, or in different orders, as long as the desired results of the technical solutions disclosed in the present disclosure can be achieved, and the present disclosure is not limited herein.
The above detailed description should not be construed as limiting the scope of the disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made in accordance with design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present disclosure should be included in the scope of protection of the present disclosure.
Claims (16)
1. A trajectory prediction method, comprising:
determining a first predicted trajectory of a target object based on a sequence of dynamic features and static features associated with the target object;
determining a second predicted trajectory of the target object based on the historical trajectory of the target object; and
determining a target predicted trajectory of the target object based on the first predicted trajectory and the second predicted trajectory;
wherein the historical track comprises a plurality of tracks;
the determining a second predicted trajectory of the target object based on the historical trajectory of the target object comprises:
carrying out grid coding on a plurality of historical tracks to obtain respective coding results of the plurality of historical tracks;
clustering the plurality of historical tracks based on the respective coding results of the plurality of historical tracks, and determining at least one target historical track;
determining the similarity between the at least one target historical track and the existing track of the target object in the current time period; and
and determining the target historical track with the similarity greater than or equal to a preset similarity threshold as the second predicted track of the target object.
2. The method of claim 1, wherein the determining a first predicted trajectory of a target object based on a sequence of dynamic features and static features associated with the target object comprises:
inputting the dynamic characteristic sequence and the static characteristic associated with the target object into a convolution module of an encoder to obtain an abstract characteristic;
inputting the abstract features into a feature extraction module of an encoder to obtain semantic vectors; and
and inputting the semantic vector into a decoder to obtain the first predicted track.
3. The method of claim 1, wherein the determining a first predicted trajectory of a target object based on a sequence of dynamic features and static features associated with the target object comprises:
inputting the dynamic characteristic sequence and the static characteristic associated with the target object into a plurality of base learners respectively to obtain respective prediction results of the plurality of base learners; and
inputting the prediction results of the plurality of base learners into a secondary learner to obtain a first prediction track of the target object;
wherein the secondary learner and each of the plurality of base learners respectively include at least one of:
the system comprises a coding and decoding model comprising a Gaussian process regression network, a coding and decoding model comprising a gray network, a coding and decoding model comprising a multilayer perceptron, a coding and decoding model comprising a convolutional neural network and a long-short term memory network, and a coding and decoding model comprising a gating cycle unit.
4. The method of claim 1, wherein the determining a target predicted trajectory of the target object based on the first predicted trajectory and the second predicted trajectory comprises:
fusing the first predicted track and the second predicted track to determine a target predicted track of the target object;
wherein the fusion process comprises at least one of:
voting, averaging, and weighting.
5. The method of claim 1, wherein the dynamic signature sequence comprises at least one of:
speed, heading, location, weather, time, ambient environment;
wherein the static features comprise attribute features of the target object.
6. The method of claim 1, wherein the sequence of dynamic features further comprises implicit semantic features;
the method further comprises the following steps:
carrying out grid coding on the existing track of the target object to obtain a coding result;
processing the coding result to obtain a semantic vector corresponding to the existing track of the target object;
based on the semantic vector, determining implicit semantic features of the existing trajectory of the target object.
7. A collision detection method, comprising:
determining respective target predicted trajectories of a plurality of objects within a target region;
determining whether at least two intersected target prediction tracks exist in the target prediction tracks of the plurality of objects; and
determining that at least two objects corresponding to at least two intersecting target predicted trajectories are at risk of collision if it is determined that there are at least two intersecting target predicted trajectories;
wherein the target predicted trajectory is predicted by the trajectory prediction method according to claims 1 to 6.
8. A trajectory prediction device comprising:
a first determination module to determine a first predicted trajectory of a target object based on a sequence of dynamic features and static features associated with the target object;
a second determination module to determine a second predicted trajectory of the target object based on the historical trajectory of the target object; and
a third determination module for determining a target predicted trajectory of the target object based on the first predicted trajectory and the second predicted trajectory;
wherein the historical track comprises a plurality of tracks;
the second determining module includes:
the encoding unit is used for carrying out grid encoding on a plurality of historical tracks to obtain respective encoding results of the plurality of historical tracks;
the clustering unit is used for clustering the plurality of historical tracks based on the respective coding results of the plurality of historical tracks and determining at least one target historical track;
the calculation unit is used for determining the similarity between the at least one target historical track and the existing track of the target object in the current time period; and
a determining unit, configured to determine a target history trajectory of which the similarity is greater than or equal to a preset similarity threshold as the second predicted trajectory of the target object.
9. The apparatus of claim 8, wherein the first determining means comprises:
the first input unit is used for inputting the dynamic characteristic sequence and the static characteristic associated with the target object into a convolution module of an encoder to obtain abstract characteristics;
the second input unit is used for inputting the abstract features into a feature extraction module of an encoder to obtain semantic vectors; and
and the decoding unit is used for inputting the semantic vector into a decoder to obtain the first prediction track.
10. The apparatus of claim 8, wherein the first determining means comprises:
a third input unit, configured to input the dynamic feature sequence and the static feature associated with the target object into a plurality of base learners, respectively, to obtain prediction results of the plurality of base learners; and
a fourth input unit, configured to input prediction results of the plurality of basis learners into a secondary learner, so as to obtain a first predicted trajectory of the target object;
wherein the secondary learner and each of the plurality of base learners respectively include at least one of:
the system comprises a coding and decoding model comprising a Gaussian process regression network, a coding and decoding model comprising a gray network, a coding and decoding model comprising a multilayer perceptron, a coding and decoding model comprising a convolutional neural network and a long-short term memory network, and a coding and decoding model comprising a gating cycle unit.
11. The apparatus of claim 8, wherein the third determining means comprises:
the fusion unit is used for performing fusion processing on the first predicted track and the second predicted track and determining a target predicted track of the target object;
wherein the fusion process comprises at least one of:
voting, averaging, and weighting.
12. The apparatus of claim 8, wherein the dynamic signature sequence comprises at least one of:
speed, heading, location, weather, time, ambient environment;
wherein the static features comprise attribute features of the target object.
13. The apparatus of claim 8, wherein the sequence of dynamic features further comprises implicit semantic features;
the device further comprises:
the encoding module is used for carrying out grid encoding on the existing track of the target object to obtain an encoding result;
the processing module is used for processing the coding result to obtain a semantic vector corresponding to the existing track of the target object;
and the conversion module is used for determining the implicit semantic features of the existing track of the target object based on the semantic vector.
14. A collision detecting device comprising:
a fourth determining module, configured to determine respective target predicted trajectories of multiple objects within the target region;
a fifth determining module, configured to determine whether at least two intersected target predicted trajectories exist in the target predicted trajectories of the respective multiple objects; and
a sixth determining module, configured to determine that there is a risk of collision of at least two objects corresponding to at least two intersecting target predicted trajectories if it is determined that there are at least two intersecting target predicted trajectories;
wherein the target predicted trajectory is predicted by the trajectory prediction method according to claims 1 to 6.
15. An electronic device, comprising:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the trajectory prediction method of any one of claims 1-6 or the collision detection method of claim 7.
16. A non-transitory computer readable storage medium storing computer instructions for causing a computer to perform the trajectory prediction method of any one of claims 1-6 or the collision detection method of claim 7.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210838283.1A CN115131393B (en) | 2021-08-16 | 2021-08-16 | Track prediction method, collision detection device, electronic equipment and medium |
CN202110940464.0A CN113658214B (en) | 2021-08-16 | 2021-08-16 | Trajectory prediction method, collision detection method, apparatus, electronic device, and medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110940464.0A CN113658214B (en) | 2021-08-16 | 2021-08-16 | Trajectory prediction method, collision detection method, apparatus, electronic device, and medium |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210838283.1A Division CN115131393B (en) | 2021-08-16 | 2021-08-16 | Track prediction method, collision detection device, electronic equipment and medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113658214A CN113658214A (en) | 2021-11-16 |
CN113658214B true CN113658214B (en) | 2022-08-09 |
Family
ID=78479361
Family Applications (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210838283.1A Active CN115131393B (en) | 2021-08-16 | 2021-08-16 | Track prediction method, collision detection device, electronic equipment and medium |
CN202110940464.0A Active CN113658214B (en) | 2021-08-16 | 2021-08-16 | Trajectory prediction method, collision detection method, apparatus, electronic device, and medium |
Family Applications Before (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210838283.1A Active CN115131393B (en) | 2021-08-16 | 2021-08-16 | Track prediction method, collision detection device, electronic equipment and medium |
Country Status (1)
Country | Link |
---|---|
CN (2) | CN115131393B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115131393B (en) * | 2021-08-16 | 2024-04-23 | 北京百度网讯科技有限公司 | Track prediction method, collision detection device, electronic equipment and medium |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114464015B (en) * | 2022-02-08 | 2023-05-02 | 北京百度网讯科技有限公司 | Data processing method, device, electronic equipment and storage medium |
CN114595770B (en) * | 2022-03-09 | 2023-10-10 | 中国电子科技集团公司第五十四研究所 | Long time sequence prediction method for ship track |
CN114385872B (en) * | 2022-03-23 | 2022-06-03 | 中国民航科学技术研究院 | Method and device for predicting eddy current dissipation rate, electronic equipment and storage medium |
CN114722232B (en) * | 2022-04-18 | 2023-12-12 | 北京航迹科技有限公司 | Method, device, equipment and storage medium for predicting motion trail |
CN116881750B (en) * | 2023-07-05 | 2024-03-29 | 广东海洋大学 | Track clustering method |
CN117475090B (en) * | 2023-12-27 | 2024-06-11 | 粤港澳大湾区数字经济研究院(福田) | Track generation model, track generation method, track generation device, terminal and medium |
CN118658338A (en) * | 2024-08-19 | 2024-09-17 | 浙江长龙海运有限公司 | Ship collision early warning method and related equipment |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101639988A (en) * | 2009-05-15 | 2010-02-03 | 天津七一二通信广播有限公司 | Method for preventing boats from colliding |
CN111209457A (en) * | 2019-12-24 | 2020-05-29 | 中国科学院电子学研究所苏州研究院 | Target typical activity pattern deviation warning method |
CN112133089A (en) * | 2020-07-21 | 2020-12-25 | 西安交通大学 | Vehicle track prediction method, system and device based on surrounding environment and behavior intention |
CN112665590A (en) * | 2020-12-11 | 2021-04-16 | 国汽(北京)智能网联汽车研究院有限公司 | Vehicle track determination method and device, electronic equipment and computer storage medium |
US11061727B1 (en) * | 2019-08-28 | 2021-07-13 | Amazon Technologies, Inc. | Preventing performance degradation over time in a cache |
CN113240199A (en) * | 2021-06-07 | 2021-08-10 | 广西民族大学 | Port ship track prediction method based on DILATE _ TLSTM |
Family Cites Families (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107045722B (en) * | 2017-03-27 | 2019-07-30 | 西安电子科技大学 | Merge the video signal process method of static information and multidate information |
CA3128038A1 (en) * | 2019-01-31 | 2020-08-06 | Rypplzz, Inc. | Systems and methods for augmented reality with precise tracking |
US20210094569A1 (en) * | 2019-09-27 | 2021-04-01 | Honda Motor Co., Ltd. | System and method for providing accurate trajectory following for automated vehicles in dynamic environments |
WO2021134172A1 (en) * | 2019-12-30 | 2021-07-08 | 华为技术有限公司 | Trajectory prediction method and related device |
CN111474953B (en) * | 2020-03-30 | 2021-09-17 | 清华大学 | Multi-dynamic-view-angle-coordinated aerial target identification method and system |
CN111401531A (en) * | 2020-04-24 | 2020-07-10 | 中国人民解放军国防科技大学 | Trajectory prediction method and system |
CN111942407B (en) * | 2020-07-31 | 2022-09-23 | 商汤集团有限公司 | Trajectory prediction method, apparatus, device and storage medium |
CN112040413B (en) * | 2020-08-06 | 2023-06-20 | 杭州数梦工场科技有限公司 | User track calculation method and device and electronic equipment |
CN112487948B (en) * | 2020-11-27 | 2022-05-13 | 华中师范大学 | Multi-space fusion-based concentration perception method for learner in learning process |
CN112348293A (en) * | 2021-01-07 | 2021-02-09 | 北京三快在线科技有限公司 | Method and device for predicting track of obstacle |
CN112766468B (en) * | 2021-04-08 | 2021-07-30 | 北京三快在线科技有限公司 | Trajectory prediction method and device, storage medium and electronic equipment |
CN115131393B (en) * | 2021-08-16 | 2024-04-23 | 北京百度网讯科技有限公司 | Track prediction method, collision detection device, electronic equipment and medium |
-
2021
- 2021-08-16 CN CN202210838283.1A patent/CN115131393B/en active Active
- 2021-08-16 CN CN202110940464.0A patent/CN113658214B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101639988A (en) * | 2009-05-15 | 2010-02-03 | 天津七一二通信广播有限公司 | Method for preventing boats from colliding |
US11061727B1 (en) * | 2019-08-28 | 2021-07-13 | Amazon Technologies, Inc. | Preventing performance degradation over time in a cache |
CN111209457A (en) * | 2019-12-24 | 2020-05-29 | 中国科学院电子学研究所苏州研究院 | Target typical activity pattern deviation warning method |
CN112133089A (en) * | 2020-07-21 | 2020-12-25 | 西安交通大学 | Vehicle track prediction method, system and device based on surrounding environment and behavior intention |
CN112665590A (en) * | 2020-12-11 | 2021-04-16 | 国汽(北京)智能网联汽车研究院有限公司 | Vehicle track determination method and device, electronic equipment and computer storage medium |
CN113240199A (en) * | 2021-06-07 | 2021-08-10 | 广西民族大学 | Port ship track prediction method based on DILATE _ TLSTM |
Non-Patent Citations (2)
Title |
---|
Designing a prediction-based clustering algorithm for target tracking in wireless sensor networks;Fatemeh Deldar et al.;《2011 International Symposium on Computer Networks and Distributed Systems (CNDS)》;20110505;全文 * |
基于AIS的船舶轨迹分析的研究与应用;杨博辰;《中国优秀博硕士学位论文全文数据库(硕士)工程科技Ⅱ辑》;20180915;全文 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115131393B (en) * | 2021-08-16 | 2024-04-23 | 北京百度网讯科技有限公司 | Track prediction method, collision detection device, electronic equipment and medium |
Also Published As
Publication number | Publication date |
---|---|
CN115131393B (en) | 2024-04-23 |
CN115131393A (en) | 2022-09-30 |
CN113658214A (en) | 2021-11-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113658214B (en) | Trajectory prediction method, collision detection method, apparatus, electronic device, and medium | |
Xiao et al. | Maritime traffic probabilistic forecasting based on vessels’ waterway patterns and motion behaviors | |
CN115879535A (en) | Training method, device, equipment and medium for automatic driving perception model | |
Wang et al. | A novel method for ship trajectory prediction in complex scenarios based on spatio-temporal features extraction of AIS data | |
Wang et al. | Data-driven methods for detection of abnormal ship behavior: Progress and trends | |
CN115909749B (en) | Vehicle running road risk early warning method, device, equipment and storage medium | |
CN114771572A (en) | Automatic driving track prediction method, device, equipment and storage medium | |
Xiao et al. | Bidirectional data-driven trajectory prediction for intelligent maritime traffic | |
CN112445847A (en) | Ship track prediction method and device based on Beidou navigation time-frequency data | |
CN115050214A (en) | Ship collision risk prediction method based on AIS data | |
Huang et al. | Tripleconvtransformer: A deep learning vessel trajectory prediction method fusing discretized meteorological data | |
CN113836661B (en) | Time prediction method, model training method, related device and electronic equipment | |
CN116092055A (en) | Training method, acquisition method, device, equipment and automatic driving vehicle | |
Xu et al. | Improved Vessel Trajectory Prediction Model Based on Stacked‐BiGRUs | |
Donandt et al. | Short-term inland vessel trajectory prediction with encoder-decoder models | |
CN114333416A (en) | Vehicle risk early warning method and device based on neural network and automatic driving vehicle | |
Liu et al. | An online method for ship trajectory compression using AIS data | |
CN117408372A (en) | Ship track prediction method and system based on deep learning | |
Liang et al. | A survey of distance-based vessel trajectory clustering: Data pre-processing, methodologies, applications, and experimental evaluation | |
Lu et al. | A Novel CNN-LSTM Method for Ship Trajectory Prediction | |
Burger et al. | Extended linear regression model for vessel trajectory prediction with a-priori AIS information | |
CN114565170A (en) | Pollutant tracing method and device, equipment, medium and product | |
US20230186099A1 (en) | Learning device, learning method, and learning program | |
Tang et al. | An Efficientnet Based Method for Autonomous Vehicles Trajectory Prediction | |
CN112418730A (en) | Construction method of response index estimation model of transportation system |
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 |