WO2021115143A1

WO2021115143A1 - Motion trajectory processing method, medium, apparatus, and device

Info

Publication number: WO2021115143A1
Application number: PCT/CN2020/132333
Authority: WO
Inventors: 刘亘轶
Original assignee: 华为技术有限公司
Priority date: 2019-12-13
Filing date: 2020-11-27
Publication date: 2021-06-17
Also published as: CN112989220A

Abstract

The present application relates to a motion trajectory processing method, comprising: obtaining a motion trajectory set, the motion trajectory set comprising a plurality of motion trajectories; determining motion path elements according to the motion trajectory set, the motion path elements comprising at least one of a center position and an intersection point set, the center position being shared by a plurality of grids and the motion trajectory set being contained in a geographic range of a plurality of grids; the intersection set comprises a plurality of intersections related to a plurality of trajectory segment clusters, each of the plurality of trajectory segment clusters comprising a plurality of trajectory segments of a plurality of motion trajectories, and furthermore, the distance between adjacent trajectory segments among the plurality of trajectory segments being less than a predetermined value, the adjacent trajectory segments respectively belonging to different motion trajectories among the plurality of motion trajectories. In the present application, determining the center position and/or intersection set of a plurality of motion trajectories can reduce the cost of storage and processing of motion trajectories, improving motion-trajectory processing efficiency.

Description

Method, medium, device and equipment for processing motion trajectory

This application claims the priority of a Chinese patent application filed with the Chinese Patent Office on December 13, 2019, the application number is 201911283885.X, and the application name is "a method, medium, device and equipment for processing trajectories", and its entire contents Incorporated in this application by reference.

Technical field

One or more embodiments of the present application generally relate to the field of artificial intelligence, and specifically relate to a motion trajectory processing method, medium, device, and equipment.

Background technique

In the prior art, when recommending a movement route, a mobile terminal is usually used to collect multiple movement trajectory data of multiple users, and the movement trajectory data including multiple GPS coordinate points are sent to the cloud-side server through the network, and the server transmits these movement data to the cloud server. The trajectory data is directly stored as the motion route, or the center of mass calculation algorithm of the regular trajectory shape is used to store the motion trajectory data and the GPS coordinates of the center of mass as the motion route. Finally, when other users request an exercise route, they search for an exercise route close to the requesting user from the cloud-side database according to the GPS coordinates of the requesting user, and return it to the requesting user.

Summary of the invention

The following describes the application from multiple aspects, and the implementation and beneficial effects of the following multiple aspects can be referred to each other.

In a first aspect, an embodiment of the present application provides a motion trajectory processing method, the method includes: obtaining a motion trajectory set, wherein the motion trajectory set includes a plurality of motion trajectories; and determining a motion route element according to the motion trajectory set, wherein the motion route element It includes at least one of a center position and an intersection point set, where the center position is shared by multiple grids and the motion track set is contained within the geographic range of the multiple grids; wherein the intersection point set includes information related to multiple trajectory segment clusters. Multiple intersections, where each of the multiple trajectory segment clusters includes multiple trajectory segments of multiple motion trajectories, and the distance between adjacent trajectory segments in the multiple trajectory segments is less than a predetermined value, where adjacent The trajectory segments respectively belong to different motion trajectories among the multiple motion trajectories.

According to the embodiment of the present application, instead of directly storing and processing multiple motion trajectories, multiple trajectory segments of the multiple motion trajectories of the user are clustered, and multiple intersection points related to the multiple trajectory segment clusters are obtained as slaves. The elements of the running route extracted from the user's multiple motion trajectories can reduce the storage and processing cost of the motion trajectory, and improve the processing efficiency of the motion trajectory; instead of making each grid correspond to a central position, multiple grids can share one The central position can also reduce the storage and processing cost of the motion trajectory, and improve the processing efficiency of the motion trajectory.

In some embodiments, the motion route element further includes: a Geohash code corresponding to each grid in the plurality of grids, which is used to index the geographic range corresponding to each grid.

According to the embodiment of the present application, by using geohash code values of multiple grids to index the movement route, the retrieval efficiency of the movement route can be improved.

In some embodiments, Geohash encoding includes at least 7-bit Base32 encoded characters.

In some embodiments, the motion route element further includes: an intersection link table indicating a link relationship between each intersection of the plurality of intersections and at least one other intersection of the plurality of intersections.

In some embodiments, the center position includes the GPS coordinates of the center position, and each of the multiple intersection points includes relative coordinates in a relative coordinate system with the center position as the origin.

According to some embodiments of the present application, by taking the center position as the origin and performing relative coordinate conversion on each intersection of the movement route, the privacy of the user can be protected and sensitive data related to the user can be eliminated.

In some embodiments, determining the movement route elements according to the set of movement trajectories includes: selecting the four most east, south, west, and north most intersections from a plurality of intersections; and determining the geography of the multiple grids based on the four intersections Range, and determine the center point of the geographic range as the center position.

In some embodiments, determining the motion route element according to the motion trajectory set includes: dividing each motion trajectory in the motion trajectory set into a plurality of trajectory segments; and dividing a plurality of trajectories whose distance between adjacent trajectory segments is less than a predetermined value The segment cluster is a trajectory segment cluster among multiple trajectory segment clusters;

In some embodiments, according to multiple trajectory segments in each trajectory segment cluster in the multiple trajectory clusters, a cluster vector corresponding to each trajectory segment cluster is determined to form multiple cluster vectors corresponding to the multiple trajectory segment clusters ; Determine the intersection between adjacent cluster vectors among multiple cluster vectors.

In some embodiments, the start point coordinates of the cluster vector are the average value of the start point coordinates of the multiple trajectory segments in the trajectory segment cluster, and the end point coordinates of the cluster vector are the average value of the end point coordinates of the multiple trajectory segments in the trajectory segment cluster.

In a second aspect, an embodiment of the present application provides a motion trajectory processing method, which includes: obtaining multiple current motion trajectory intersection points related to a current motion trajectory set and multiple historical motion trajectory intersection points related to a historical motion trajectory set; And at least partly based on the abnormal segment deviation, determining whether to retain at least one of the multiple current motion trajectory intersections and/or at least one of the multiple historical motion trajectory intersections, so as to obtain an optimized motion route , Wherein the abnormal segment deviation at least partially specifies the proportion of abnormal trajectory segments in the multiple trajectory segments in the current trajectory set, where the distance from the abnormal trajectory segment to at least one class center of the multiple trajectory segments exceeds the first predetermined value.

According to the embodiment of the present application, when fusing the current motion trajectory set and the historical motion trajectory set, the abnormal segment deviation of the current motion trajectory set is considered, and the abnormal segment deviation is the same as the GPS when multiple motion trajectories in the current motion trajectory set are generated. Signal quality is related, so the abnormal segment deviation of the current motion trajectory set is used to determine whether to retain at least one current motion trajectory intersection among multiple current motion trajectory intersections and/or at least one historical motion trajectory intersection among multiple historical motion trajectory intersections. More accurately controlling the fusion error can also improve the fusion quality.

In some embodiments, the current motion trajectory set and the historical motion trajectory set are contained in the same geographic scope.

In some embodiments, determining whether to retain at least one current motion trajectory intersection point among the multiple current motion trajectory intersection points and/or at least one historical motion trajectory intersection point among the multiple historical motion trajectory intersection points includes using LSTM (Long short-term memory , Long-short-term memory) model determines whether to retain at least one current motion trajectory intersection among multiple current motion trajectory intersections and/or at least one historical motion trajectory intersection among multiple historical motion trajectory intersections.

In some embodiments, determining whether to retain at least one current motion trajectory intersection point among the multiple current motion trajectory intersection points and/or at least one historical motion trajectory intersection point among the multiple historical motion trajectory intersection points further includes: using the LSTM model according to the following Formula calculation is used to determine whether to retain at least one of the multiple historical motion trajectory intersection points of the forgetting gate:

f _t =σ(W _f ·[h _t-1 ,X _t ]+b _f )

b _f =1+b _m

Among them, f _t represents the forgetting gate of the neuron at the t-th moment of the LSTM model, σ represents the sigmoid function, W _f represents the weight matrix of the forgetting gate, and X _t represents the input of the neuron at the t-th moment of the LSTM model and includes Multiple current motion trajectory intersections, h _t-1 represents the output of the neuron at the t-1 time of the LSTM model and includes multiple historical motion trajectory intersections, [h _t-1 ,X _t ] represents h _t-1 The long vector formed by splicing with X _t _{, b f} represents the offset value of the forgetting gate, and b _m represents the deviation of the abnormal segment.

In some embodiments, determining whether to retain at least one current motion trajectory intersection of the multiple current motion trajectory intersections and/or at least one historical motion trajectory intersection of the multiple historical motion trajectory intersections further includes using an LSTM model according to the following formula Calculate the input gate used to determine whether to keep at least one of the multiple current motion trajectory intersection points:

i _t =σ(W _i ·[h _t-1 ,X _t ]+b _i )

Wherein, i _t represents neurons t-th time LSTM model input gate, σ represents a sigmoid function, W _i represents the weight of the input gate of the weight matrix, X _t denotes an input neuron t-th time LSTM models and comprising Multiple current motion trajectory intersections, h _t-1 represents the output of the neuron at the t-1 time of the LSTM model and includes multiple historical motion trajectory intersections, [h _t-1 ,X _t ] represents h _t-1 The long vector formed by splicing with X _t _{, b i} represents the bias value of the input gate.

In some embodiments, determining whether to retain at least one current motion trajectory intersection of the multiple current motion trajectory intersections and/or at least one historical motion trajectory intersection of the multiple historical motion trajectory intersections further includes using an LSTM model according to the following formula It is determined whether to output at least one current motion trajectory intersection point among the multiple current motion trajectory intersection points and at least one historical motion trajectory intersection point among the multiple historical motion trajectory intersection points, so as to obtain an optimized motion route:

o _t =σ(W _o ·[h _t-1 ,X _t ]+b _o )

h _t ＝o _t ·tanh(C _t )

Among them, X _t represents the input of the neuron at the t-th time of the LSTM model and includes multiple current trajectory intersections, and h _t-1 represents the output of the neuron at the t-1th time of the LSTM model and includes multiple histories Point of intersection of motion trajectories,

Represents the input cell state of the neuron at the t-th time of the _{LSTM model, W c} represents the weight matrix of the input cell state of _{the neuron at the t-th time of the LSTM model, and b c} represents the neuron at the t-th time of the LSTM model The input cell state bias value, C _t represents the cell state of the neuron at the t time of the LSTM model, C _t-1 represents the cell state of the neuron at the t-1 time of the LSTM model, and f _t represents The forget gate of the neuron at the t-th time of the _{LSTM model, i t} represents the input gate of the neuron at the t-th time of the _{LSTM model, o t} represents the output gate of the neuron at the t-th time of the LSTM model, W _o Represents the weight matrix of the output gate, b _o represents the bias value of the output gate, and h _t represents the output of the neuron at the t-th time of the LSTM model and may include multiple intersection points of the optimized movement route.

In some embodiments, the method further includes selecting the four most eastward, most southerly, most westward, and most northerly intersection points from the multiple intersection points of the optimized motion route;

According to the four intersection points, the geographic range containing the optimized motion route is determined, and the center point of the geographic range is determined as the center position of the optimized motion route.

In some embodiments, the method further includes determining a plurality of grids of a geographic range and geohash coding values of the plurality of grids, wherein the plurality of grids share a central location.

According to the embodiment of the present application, instead of making each grid correspond to a central position and making multiple grids share a central position, the storage and processing cost of the optimized movement route can be reduced, and the processing efficiency of the optimized movement route can be improved; In addition, by indexing the optimized movement route using geohash code values of multiple grids, the retrieval efficiency of the optimized movement route can be improved.

In some embodiments, the center position includes the GPS coordinates of the center position, and each of the multiple intersection points of the optimized movement route includes relative coordinates in a relative coordinate system with the center position as the origin.

According to the embodiment of the present application, by taking the center position as the origin and performing relative coordinate conversion on each intersection of the optimized movement route, the privacy of the user can be protected and sensitive data related to the user can be eliminated.

In some embodiments, the method further includes: clustering the centers of the plurality of trajectory segments and obtaining at least one cluster center of the plurality of trajectory segments.

In the third aspect, an embodiment of the present application provides a method for recommending a sports route. The method includes: receiving a user's recommendation request for a sports route; acquiring the user's GPS coordinates, and calculating the user's GPS (Global Positioning System, global positioning system). ) The geohash code value of the coordinate; from multiple grid geohash code values, query the grid geohash code value with the smallest difference with the user's GPS coordinate geohash code value; obtain the smallest difference with the user's GPS coordinate geohash code value The grid geohash code value corresponds to the motion route and the center position of the motion route, where the center position is shared by multiple grids containing the geographic range of the motion route; and the motion route is recommended to the user according to the center position.

According to the embodiment of the present application, instead of making each grid correspond to a central position and making multiple grids share a central position, the storage and processing cost of the movement route can be reduced, and the processing efficiency of the movement route can be improved; in addition, by using The geohash code values of multiple grids index the movement route, which can improve the retrieval efficiency of the movement route.

In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium that stores instructions on the computer-readable storage, and when the instructions run on a machine, the machine executes any of the above methods.

In a fifth aspect, an embodiment of the present application provides a motion route processing device, which includes: a processor; a memory, which stores instructions on the memory, and when the instructions are executed by the processor, the processor executes any of the above methods .

In a sixth aspect, an embodiment of the present application provides a motion trajectory processing device. The device includes a motion trajectory recording module for acquiring a motion trajectory set, where the motion trajectory set includes multiple motion trajectories; and a motion route extraction module for Determine the motion route element according to the motion trajectory set, where the motion route element includes at least one of a center position and an intersection set, wherein the center position is shared by multiple grids and the motion track set is contained in the geographic range of the multiple grids; And wherein, the intersection set includes multiple intersections related to multiple trajectory segment clusters, wherein each trajectory segment cluster in the multiple trajectory segment clusters includes multiple trajectory segments of multiple motion trajectories, and the phases in the multiple trajectory segments The distance between adjacent track segments is less than a predetermined value, and the adjacent track segments respectively belong to different motion tracks among the multiple motion tracks.

According to the embodiment of the present application, instead of directly storing and processing multiple motion trajectories, multiple trajectory segments of the multiple motion trajectories of the user are clustered, and multiple intersection points related to the multiple trajectory segment clusters are obtained as the user The multiple motion trajectory elements can reduce the storage and processing cost of the motion trajectory, and improve the processing efficiency of the motion trajectory; instead of making each grid correspond to a central position, making multiple grids share a central position can also reduce The storage and processing cost of the motion trajectory improves the processing efficiency of the motion trajectory.

In some embodiments, the motion route extraction module is further used to determine the relative coordinates of each of the multiple intersection points in a relative coordinate system with the center position as the origin.

In some embodiments, the movement route extraction module is used to determine the movement route elements according to the set of movement trajectories, including: selecting the four most east, south, west, and northest intersections from a plurality of intersections; according to the four The intersection point determines the geographic range of multiple grids, and determines the center point of the geographic range as the center position.

In some embodiments, the motion route extraction module is used to determine the motion route elements according to the motion trajectory set, including: dividing each motion trajectory in the motion trajectory set into multiple trajectory segments; The multiple trajectory segments whose spacing is less than the predetermined value are clustered into one trajectory segment cluster in the multiple trajectory segment clusters; according to the multiple trajectory segments in each trajectory segment cluster in the multiple trajectory clusters, it is determined to correspond to each track segment The cluster vectors of the clusters thus form multiple cluster vectors corresponding to multiple trajectory segment clusters; the intersection between adjacent cluster vectors in the multiple cluster vectors is determined.

In a seventh aspect, an embodiment of the present application provides a motion trajectory processing device, which includes: a motion route extraction module, configured to obtain multiple current motion trajectory intersections related to the current motion trajectory set and information related to the historical motion trajectory set Multiple historical motion trajectory intersections; and a motion route optimization module that determines whether to retain at least one of the multiple current motion trajectory intersections and/or at least one of the multiple historical motion trajectory intersections based at least in part on the abnormal segment deviation A historical motion trajectory intersection point to obtain an optimized motion route; where the abnormal segment deviation at least partially specifies the proportion of abnormal trajectory segments in the multiple trajectory segments of the current motion trajectory set, where the distance of the abnormal trajectory segment is large At least one class center of the trajectory segments exceeds the first predetermined value.

According to the embodiment of the present application, when fusing the current motion trajectory set and the historical motion trajectory set, the abnormal segment deviation of the current motion trajectory set is considered, and the abnormal segment deviation is the same as the GPS when multiple motion trajectories in the current motion trajectory set are generated. The signal quality is related, so the abnormal segment deviation of the current motion trajectory set is used to determine whether to discard multiple historical motion trajectory intersections, which can more accurately control the fusion error and improve the fusion quality.

In some embodiments, the motion route optimization module is used to determine whether to retain at least one current motion trajectory intersection point among the multiple current motion trajectory intersection points and/or at least one historical motion trajectory intersection point among the multiple historical motion trajectory intersection points, including: A LSTM (Long short-term memory) model is used to determine whether to retain at least one current motion trajectory intersection among multiple current motion trajectory intersections and/or at least one historical motion trajectory intersection among multiple historical motion trajectory intersections.

In some embodiments, the motion route optimization module is used to determine whether to retain at least one current motion trajectory intersection among the multiple current motion trajectory intersections and/or at least one historical motion trajectory intersection among the multiple historical motion trajectory intersections. Yu: Using the LSTM model, calculate the forgetting gate used to determine whether to retain at least one historical motion trajectory intersection point among multiple historical motion trajectory intersection points according to the following formula:

f _t =σ(W _f ·[h _t-1 ,X _t ]+b _f )

b _f =1+b _m

Among them, f _t represents the forgetting gate of the neuron at the t-th moment of the LSTM model, σ represents the sigmoid function, W _f represents the weight matrix of the forgetting gate, and X _t represents the input of the neuron at the t-th moment of the LSTM model and includes Multiple current motion trajectory intersections, h _t-1 represents the output of the neuron at the t-1 time of the LSTM model and includes multiple historical motion trajectory intersections, [h _t-1 ,X _t ] represents the combination of h _tt and X The long vector formed by splicing _t _{, b f} represents the offset value of the forget gate, and b _m represents the deviation of the abnormal segment.

In some embodiments, the motion route optimization module is used to determine whether to retain at least one current motion trajectory intersection among the multiple current motion trajectory intersections and/or at least one historical motion trajectory intersection among the multiple historical motion trajectory intersections. Yu: Using the LSTM model, calculate the input gate used to determine whether to retain at least one of the current motion trajectory intersection points according to the following formula:

i _t =σ(W _i ·[h _t-1 ,X _t ]+b _i )

In some embodiments, the motion route optimization module is used to determine whether to retain at least one current motion trajectory intersection among the multiple current motion trajectory intersections and/or at least one historical motion trajectory intersection among the multiple historical motion trajectory intersections. Yu: Using the LSTM model, determine whether to output at least one of the multiple current motion trajectory intersections and at least one of the multiple historical motion trajectory intersections according to the following formula, so as to obtain an optimized motion route:

o _t =σ(W _o ·[h _t-1 ,X _t ]+b _o )

h _t ＝o _t ·tanh(C _t )

In some embodiments, the motion route extraction module is also used to select the four most east, south, west, and northest intersections from the multiple intersections of the optimized motion route; determine the included movement route according to the four intersections The geographic range of the geographic area is determined, and the center point of the geographic range is determined as the center position of the optimized movement route.

In some embodiments, the motion route extraction module is also used to determine multiple grids in a geographic range and geohash coding values of multiple grids, where multiple grids share a central position.

In some embodiments, the motion route extraction module is further used to determine the relative coordinates of each of the multiple intersection points of the optimized motion route in the relative coordinate system with the center position as the origin.

According to the embodiment of the present application, by taking the center position as the origin and performing the relative coordinate conversion on each intersection of the movement route, the privacy of the user can be protected and the sensitive data related to the user can be eliminated.

In some embodiments, the motion route extraction module is also used to:

Cluster the centers of the multiple trajectory segments and obtain at least one cluster center of the multiple trajectory segments.

Description of the drawings

FIG. 1 shows a schematic diagram of a scene in which a user's motion track is recorded through a terminal device according to an embodiment of the present application;

FIG. 2 shows a schematic structural diagram of a terminal device 100 according to an embodiment of the present application;

Fig. 3 shows a schematic diagram of trajectory segmentation and clustering during motion route extraction according to an embodiment of the present application;

FIG. 4 shows a schematic diagram of a doubly linked list of intersection points of movement routes according to an embodiment of the present application;

FIG. 5 shows a schematic diagram of geohash encoding of a movement route according to an embodiment of the present application;

FIG. 6 shows a schematic diagram of a model of a long and short-term memory network according to an embodiment of the present application;

FIG. 7 shows a schematic flowchart of a method for extracting a movement route according to an embodiment of the present application;

FIG. 8 shows a schematic flowchart of a method for optimizing a movement route according to an embodiment of the present application;

FIG. 9 shows a schematic flowchart of a method for recommending a sports route according to an embodiment of the present application;

Fig. 10 shows a schematic structural diagram of a motion route processing device according to an embodiment of the present application.

Detailed ways

The application will be further described below in conjunction with specific embodiments and drawings. The specific embodiments described here are only for explaining the application, rather than limiting the application. In addition, for ease of description, the drawings only show a part of the structure or process related to the present application instead of all. It should be noted that in this specification, similar reference numerals and letters indicate similar items in the following drawings. Therefore, once a certain item is defined in one drawing, it is not necessary to refer to it in subsequent drawings. To further define and explain.

FIG. 1 shows a schematic diagram of a scene of recording a user's motion track through a terminal device according to an embodiment of the present application. The terminal device, such as the terminal device 100 in FIG. 2, may include, but is not limited to, a portable or mobile device , Mobile phones, personal digital assistants, cellular phones, handheld PCs, portable media players, handheld devices, wearable devices (for example, watches, bracelets, display glasses or goggles, head-mounted displays, head-mounted devices), navigation devices , Servers, network devices, graphics devices, video game devices, set-top boxes, laptop devices, virtual reality and/or augmented reality devices, IoT devices, industrial control devices, in-vehicle infotainment devices, streaming media client devices, e-books , Reading equipment, POS machines and other equipment. As shown in Figure 1, during a user's exercise (such as, but not limited to, running, brisk walking, etc.) along a sports field (such as, but not limited to, a playground, park, community, square, road, or other venues), the terminal device, such as The terminal device 100 in FIG. 2 can determine the user's GPS (global positioning system, global positioning system) coordinates in real time according to the received satellite signals, so as to record the user's movement trajectory in real time.

FIG. 2 shows a schematic structural diagram of a terminal device 100 according to an embodiment of the present application. As shown in FIG. 2, the terminal device 100 may include, but is not limited to, a GPS positioning module 110, a motion track recording module 120, and a data center 130. , The exercise route extraction module 140, the exercise route optimization module 150, and the exercise route recommendation module 160. Among them, one or more components of the terminal device 100 (for example, one or more of the GPS positioning module 110, the movement track recording module 120, the data center 130, the movement route extraction module 140, the movement route optimization module 150, and the movement route recommendation module 160 Multiple), can be composed of application-specific integrated circuits (ASIC), electronic circuits, (shared, dedicated or group) processors and/or memories that execute one or more software or firmware programs, combinational logic circuits, and devices that provide the described functions Any combination of other suitable components. According to one aspect, the processor may be a microprocessor, a digital signal processor, a microcontroller, etc., and/or any combination thereof. According to another aspect, the processor may be a single-core processor, a multi-core processor, etc., and/or any combination thereof.

It should be noted that although FIG. 2 shows that the exercise route extraction module 140, the exercise route optimization module 150, and the exercise route recommendation module 160 are all located on the terminal device 100 on the user side, the exercise route extraction module 140 and the exercise route optimization module 150 And one or more of the exercise route recommendation module 160 may be located on other devices, such as, but not limited to, a cloud-side server. Therefore, in the embodiment of the present application, the terminal device 100 and the cloud-side server, etc. Collectively referred to as motion trajectory processing equipment. In addition, although FIG. 2 shows that the movement track recording module 120, the movement route extraction module 140, the movement route optimization module 150, and the movement route recommendation module 160 are independent modules, the movement route extraction module 140, the movement route optimization module 150, and the One or more of the exercise route recommendation modules 160 may also be used as a part of the exercise track recording module 120.

According to some embodiments of the present application, the GPS positioning module 110 is used to determine the GPS coordinates of the user (or the terminal device 100) according to the received satellite signals, and the motion track recording module 120 is used to record the user's motion track during the user's motion, specifically The sports APP can obtain the user's GPS coordinates from the GPS positioning module 110 at a certain frequency (ie, dot frequency). The GPS coordinates and the corresponding time will constitute a track point, and a motion track may include multiple track points.

According to some embodiments of the present application, the motion trajectory recorded by the motion trajectory recording module 120 may be stored in the data center 130. The data center 130 may include, but is not limited to, non-transient objects manufactured or formed by machines or equipment. A tangible arrangement of state, which includes storage media, such as hard disks or any other types of disks, including floppy disks, optical disks, compact disk read-only memory (CD-ROM), compact disk rewritable (CD-RW), and magneto-optical disks ; Semiconductor devices, such as read-only memory (ROM), random access memory (RAM) such as dynamic random access memory (DRAM) and static random access memory (SRAM), erasable programmable read-only memory ( EPROM), flash memory, electrically erasable programmable read-only memory (EEPROM); phase change memory (PCM); magnetic or optical card; or any other type of medium suitable for storing data.

According to some embodiments of the present application, the exercise route extraction module 140 is used to record the user from the exercise track recording module 120 within a period of time (for example, one day, three days, one week, two weeks, one month or other Time period) Extract a simplified motion route from multiple motion trajectories generated in the same geographic range (for example, the same sports field). Examples of multiple motion trajectories may be, but are not limited to, motion trajectories generated by the user today, yesterday, and the day before yesterday, or multiple motion trajectories obtained by dividing the motion trajectory generated by the user in one exercise, for example, The motion trajectory generated by the user running multiple laps along the sports field is divided into multiple motion trajectories obtained by dividing each lap. Wherein, the motion route model may include, but is not limited to, a set of multiple intersection points constituting the motion route, and the intersection point set includes multiple intersection points related to multiple trajectory segment clusters, wherein each of the multiple trajectory segment clusters The trajectory segment cluster includes multiple trajectory segments of the motion trajectory, and the distance between adjacent trajectory segments in the multiple trajectory segments is less than a predetermined value, wherein the adjacent trajectory segments respectively belong to different running trajectories in the multiple running trajectories .

According to some embodiments of the present application, the exercise route optimization module 150 is used to merge the intersection of the current exercise route and at least one historical exercise route in the same geographic area (for example, the same sports venue) based on the cyclic neural network 151 to obtain An optimized exercise route in the geographic range is stored in the data center 130 as a default exercise route in the geographic range. Wherein, the current exercise route refers to the exercise route extraction module 140 extracted from the current exercise trajectory set generated by the user in the current time period (for example, one day, three days, one week, two weeks, one month, or other time periods). Sports routes, each historical exercise route in at least one historical exercise route refers to the historical time period (for example, one day, three days, one week, two weeks, one month) from the user before the current time period through the exercise route extraction module 140 Or other time periods). Among them, the recurrent neural network 151 is a type of deep learning network with memory ability. The output of a neuron of the recurrent neural network 151 is not only related to the input of the neuron, but also related to the output of the previous neuron. The recurrent neural network 151 Examples of may include, but are not limited to, an LSTM (long short-term memory) network and a GRU (gated recurrent unit, gated recurrent unit) network. In one example, each neuron of the recurrent neural network 151 may Related to the movement route at different times. Wherein, the fusion includes using the cyclic neural network 151 to determine which intersections of the current movement route and each historical movement route can be retained to form an optimized movement route. Therefore, the optimized movement route includes the current movement route and one or the other of each historical movement route. Multiple intersections. Among them, the current exercise route generation user and the historical exercise route generation users may be the same or different. In different situations, the terminal device 100 may obtain the exercises of other users in the same geographic area from the server, for example, through the network. route.

According to some embodiments of the present application, the sports route recommendation module 160 is configured to receive a user's recommendation request for a sports route, obtain the user's GPS coordinates, and recommend a sports route close to the user according to the user's GPS coordinates.

The following will further describe the functions of the exercise route extraction module 140, the exercise route optimization module 150, and the exercise route recommendation module 160 of the terminal device 100 with reference to FIG. 2.

As shown in FIG. 2, the motion route extraction module 140 may include, but is not limited to, a trajectory segmentation unit 141, a trajectory clustering unit 142, a trajectory extraction unit 143, and a model element determination unit 144. Among them, one or more components of the motion route extraction module 140 (for example, one or more of the trajectory segmentation unit 141, the trajectory clustering unit 142, the trajectory extraction unit 143, and the model element determination unit 144) can be integrated by dedicated Any combination of circuits (ASIC), electronic circuits, (shared, dedicated or group) processors and/or memories that execute one or more software or firmware programs, combinational logic circuits, and other suitable components that provide the described functions . According to one aspect, the processor may be a microprocessor, a digital signal processor, a microcontroller, etc., and/or any combination thereof. According to another aspect, the processor may be a single-core processor, a multi-core processor, etc., and/or any combination thereof.

According to some embodiments of the present application, the trajectory segmentation unit 141 may be used to obtain multiple motion trajectories recorded by the motion trajectory recording module 120 from the data center 130, and divide each motion trajectory into multiple trajectory segments.

Specifically, each trajectory can be expressed as

Among them, TR _m represents the m-th motion trajectory, n _m represents the number of trajectory points included in the m-th motion trajectory,

Represents the i-th trajectory point of the m-th motion trajectory,

Represents the GPS coordinates of the i-th trajectory point of the m-th motion trajectory,

Represents the time corresponding to the GPS coordinates of the i-th trajectory point of the m-th motion trajectory.

For each motion trajectory, the trajectory dividing unit 141 may represent two trajectory points that are adjacent in time as a trajectory segment, wherein the direction of the trajectory segment is determined by the earlier trajectory point (ie, the starting trajectory point). Point to a later track point (ie, end track point), and the size of the track segment is the distance between two track points. Figure 3 shows an example of three motion trajectories TR ₁ to TR ₃ , it should be noted that only a part of the motion trajectories TR ₁ to TR ₃ are shown in the figure, where each motion trajectory is divided into multiple trajectories segment.

According to some embodiments of the present application, the trajectory clustering unit 142 may sequentially cluster the trajectory segments of each motion trajectory into a plurality of trajectory segment clusters. Among them, in a trajectory segment cluster, the trajectory segments between adjacent trajectory segments The distance is less than a predetermined value, wherein the adjacent track segments respectively belong to different running tracks in the plurality of running tracks. As an example, two trajectory segments are represented as ts _i and ts _j , where the starting point of the trajectory segment ts _i _{is si} , the end point is e _i , the starting point of the trajectory segment ts _j _{is s j} , and the end point is e _j , then The distance between _{trajectory segments ts i} and ts _j can be calculated based on the following formula:

d _|| = MIN(l _||1 , l _||2 ) Formula 3

_Wherein, l ⊥1 _J represents a path section starting point ts to the track segment s _j _i is the vertical distance _ts, l ⊥2 represents the end of the track period ts _J e _j to the track segment _i is the vertical distance ts, l _{|| 1} represents a locus _I starting period ts in the track section s _i ts s _j _J starting from the projection point in the track period ts of _I, l _{|| 2} indicates the end of the track segment e _{_i i} ts to ts track segment endpoint e _j _J The distance of the projection point on the trajectory segment ts _i , d represents the distance between the trajectory segment ts _i and ts _j .

For example, in the example shown in FIG. 3, the trajectory clustering unit 142 of FIG. 2 may first determine the first trajectory segment cluster and the reference trajectory segment of the first trajectory segment cluster, and the trajectory clustering unit 142 of FIG. Determine the trajectory segment that can be clustered with the reference trajectory segment among the trajectory segments of each motion trajectory. In an example, the trajectory clustering unit 142 of FIG. 2 may determine the reference trajectory segment of the first trajectory segment cluster as _{the trajectory segment P11 of TR 1} , and determine that it can be compatible with the trajectory segment P11 of _{TR 1 in other trajectory segments.} To cluster the trajectory segments, specifically, the trajectory clustering unit 142 of FIG. 2 may first determine that in the motion trajectories TR ₂ to TR ₃ , the distance between the trajectory segments P31 and P32 and _{the trajectory segment P11 of TR 1} is less than a predetermined value. Thus as with the track TR P11 segment ₁ and segment adjacent tracks TR track segment P11 cluster _1; FIG track clustering unit 1422 may continue to determine the track segment of P21 TR ₂ and TR track segment ₃ and P31 the spacing between the predetermined value smaller than P32 and thus can be used as the track TR ₃ segments P31 and P32 adjacent tracks TR track segment clustering section ₃ P31 and P32, then finally, the clustering unit 142 may track _{TR. 1} trajectory segment P11, TR track segment P31 and P32 ₃ and cluster TR ₂ P21 path sections for the first track segment cluster.

In another example, the trajectory clustering unit 142 of FIG. 2 may determine the reference trajectory segment of the first trajectory segment cluster as _{the trajectory segment P31 of TR 3} , and it can be matched with the trajectory segment of _{TR 3 in other trajectory segments.} P31 cluster path sections, in particular, the trajectory clustering unit may determine the track segment P11 142 tracks TR ₁ and TR ₂ and the segment of P21 spacing between the tracks TR ₃ is less than the predetermined value P31 segment, it is possible to TR ₁ TR trajectory segment P11 and P21 of a track section ₂ and section ₃ of the track TR P31 adjacent tracks TR track segment P31 clustering section ₃ for the first track segment cluster. It should be noted that the trajectory clustering unit 142 in FIG. 2 may also determine other trajectory segments as the reference trajectory segment of the first trajectory segment cluster.

Then, the trajectory clustering unit 142 may determine the second trajectory segment cluster and the reference trajectory segment of the second trajectory segment cluster. In an example, the trajectory clustering unit 142 of FIG. 2 may determine the reference trajectory segment of the second trajectory segment cluster as _{the trajectory segment P12 of TR 1} , and determine that it can be compatible with the trajectory segment P12 of _{TR 1 in other trajectory segments.} cluster path section, in particular, FIG. 2 locus clustering unit 142 can determine the distance between the track segments P33 TR ₃ and TR track segment ₁ is less than the predetermined value P12, and thus may be used as the trajectory of _a segment P12 TR adjacent tracks segment P12 of the track segment _{TR. 1} cluster, and the path section ₂ P22 TR spacing between the tracks TR section ₃ is smaller than the predetermined value P33, and thus may be used as the adjacent tracks TR section ₃ of P33 P33 track TR track segment clustering section _3, then the final trajectory clustering unit 142 may track _{TR. 1} segment of P12, P33 TR track segment ₃ and segment track TR ₂ P22 cluster for the second trail segment cluster. It should be noted that the trajectory clustering unit 142 of FIG. 2 may also determine other trajectory segments as the reference trajectory segment of the second trajectory segment cluster.

By analogy, the trajectory clustering unit 142 of FIG. 2 can determine other trajectory segment clusters of the motion trajectory. It should be noted that, for the motion trajectory that is different from that shown in FIG. 3, the trajectory segment clusters of the motion trajectory can be determined based on the principle similar to the foregoing.

According to some embodiments of the present application, for each trajectory segment cluster, the trajectory extraction unit 143 may calculate a cluster vector based on each trajectory segment in the cluster. As an example, the trajectory extraction unit 143 may use the average value of the GPS coordinates of the starting trajectory point of each trajectory segment in the trajectory segment cluster as the GPS coordinates of the starting point of the cluster vector, and the GPS coordinates of the ending trajectory point of each trajectory segment The mean value is used as the GPS coordinates of the end point of the cluster vector. As another example, the trajectory extraction unit 143 may use any trajectory segment in the trajectory segment cluster as the cluster vector. It should be noted that the cluster vector can also be determined based on any other suitable method.

According to some embodiments of the present application, the trajectory clustering unit 142 is also used to determine the abnormal trajectory segment and determine the proportion of the abnormal trajectory segment in all the trajectory segments, that is, the deviation of the abnormal segment. Specifically, the trajectory clustering unit 142 may use a hierarchical clustering algorithm, a K-means clustering algorithm or other clustering algorithms to cluster the center points of each trajectory segment of the multiple running trajectories, and obtain multiple trajectory segment center points. The trajectory clustering unit 142 can calculate the distance between the center point of each trajectory segment of multiple running trajectories and the center of each category, and the minimum distance between the center point and the center of each category is greater than a predetermined A trajectory segment with a value (the predetermined value is the same as or different from the aforementioned predetermined value related to the distance between trajectory segments) is marked as an abnormal trajectory segment. For example, in the example of FIG. 3, the trajectory segment P15 of _{TR 1} _{and the trajectory segment P35 of TR 3} may be marked as abnormal trajectory segments. In some cases, a higher abnormal segment deviation can indicate poor GPS signal quality during movement.

According to some embodiments of the present application, the model element determination unit 144 may be used to determine the model elements of the simplified motion route extracted from the motion trajectory according to the motion route model. Specifically, the motion route model may include a set of intersection points of adjacent cluster vectors of adjacent trajectory segment clusters. As an example, in the case where the adjacent cluster vectors of adjacent trajectory segment clusters do not meet end to end, the model element determining unit 144 may extend one or two of the adjacent cluster vectors to determine the adjacent cluster vector The GPS coordinates of the intersection between the adjacent cluster vectors; in the case where the adjacent cluster vectors have intersected, the model element determination unit 144 can directly determine the GPS coordinates of the intersection between the adjacent cluster vectors. The sports route model can also include the starting point of the sports route and the end of the sports route. The model element determining unit 144 may use the start point of the cluster vector of the first trajectory segment cluster of the motion trajectory as the starting point of the motion route, and the end point of the cluster vector of the last trajectory segment cluster as the end point of the motion route. The motion route model may further include an intersection link table indicating the link relationship between each intersection point in the intersection point set of the cluster vector and the other at least one intersection point, wherein the intersection point set includes the start point and the end point of the motion route. As an example, the model element determining unit 144 may determine the doubly linked list between the intersection points as shown in FIG. 4, where the starting point may be the starting point of the motion route, and the intersection point 1 may be the cluster of the first trajectory segment cluster of the motion trajectory. The intersection point between the vector and the cluster vector of the second trajectory segment cluster. Intersection point 2 can be the intersection point between the cluster vector of the second trajectory segment cluster of the motion trajectory and the cluster vector of the third trajectory segment cluster, and so on, The end point can be the end point of the movement route. In addition, each intersection shown in FIG. 4 may include the GPS coordinates of the intersection, a link to the previous intersection (for example, but not limited to, a pointer), and a link to the next intersection (for example, but not limited to, a pointer) , And the starting point may include the GPS coordinates of the starting point and a link to the next point of intersection (for example, but not limited to, a pointer), and the end point may include the GPS coordinates of the ending point, and a link to the previous point of intersection (for example, but not limited to, a pointer). The doubly linked list shown in FIG. 4 enables the user to move along the forward direction of the movement route (for example, from the starting point shown in FIG. 4 to the end point shown in FIG. 4), or along the reverse direction of the movement route (for example, from the figure The end point shown in 4 points to the starting point shown in Fig. 4) to move. It should be noted that the model element determining unit 144 may also determine a one-way linked list between intersections, that is, each intersection shown in FIG. 4 may include the GPS coordinates of the intersection and the link to the previous intersection (for example, but not Limited to pointers) or links to the next intersection (for example, but not limited to pointers).

In order to improve the retrieval efficiency of the sports route when recommending the sports route to the user, the sports route model can also include the index of the sports route, that is, the geohash code value of multiple grids representing the geographic range of the sports route, where the geohash code is An address encoding that can encode two-dimensional latitude and longitude coordinates into a one-dimensional string, for example, a base32 encoded string to improve the retrieval efficiency of geographic locations. Figure 5 shows four grids representing the geographic range of the movement route and their geohash code values. As shown in the figure, the model element determination unit 144 can determine the smallest bounding rectangle of the movement route, wherein two of the smallest bounding rectangles One edge is parallel to the line of latitude and is determined by the latitude coordinates of the southernmost and northernmost intersection points in the set of intersection points of the movement route, and the other two edges are parallel to the meridian and are determined by the longitude coordinates of the easternmost and westernmost intersection points in the set of intersection points of the movement route. In addition, the model element determining unit 144 may use the center point of the minimum circumscribed rectangle of the motion route as the center position of the motion route. For example, the GPS coordinates of the center position of the motion route shown in the figure are (121.623, 29.855).

As shown in FIG. 5, the model element determination unit 144 of FIG. 2 can take the center position as the center, divide the minimum bounding rectangle of the motion route into four grids, and determine the geohash code value corresponding to the geographic range of each grid. The number of bits of the geohash encoding value may be, for example, but not limited to, 7-bit base32 encoding. For example, as shown in FIG. 5, the encoding values of the four grids are wx4g340, wx4g341, wx4g342, and wx4g343, respectively.

It should be noted that Fig. 5 is only an example of a grid that characterizes the geographic range of the movement route. The movement route formed by each intersection is not limited to the movement route shown in Fig. 5, and the GPS coordinates of the center position are not limited to the figure. GPS coordinates shown.

In addition, in order to protect the privacy of the user, the sports route model may also include the GPS coordinates of the center position of the sports route, and the coordinates of each intersection point of the sports route are relative coordinates with respect to the center position. As an example, the model element determination unit 144 may use the aforementioned center position as the origin to construct a relative coordinate system, and convert the absolute GPS coordinates (for example, longitude and latitude) of each intersection of the movement route into relative coordinates in the relative coordinate system. Coordinates, in this way, can eliminate sensitive data related to users.

The model element determination unit 144 may also save each element of the simplified motion route extracted from the motion trajectory determined according to the motion route model in the data center 130 as a candidate motion route to be recommended to the user.

According to some embodiments of the present application, in order to further improve the quality of the exercise route recommended to the user, the exercise route optimization module 150 may be based on the recurrent neural network 151 to compare the current exercise route in the same geographic area (for example, the same sports venue) and The intersections of at least one historical movement route are merged to obtain an optimized movement route within the geographic range. Specifically, taking the recurrent neural network 151 as the LSTM network as an example to illustrate the fusion process of the current movement route and at least one historical movement route, FIG. 6 shows a schematic diagram of a model of the LSTM network. As shown in FIG. 6, the LSTM network may include a plurality of neurons a ₁ ~ a _t, each neuron associated with a movement route, e.g., neuron a ₁ associated with the initial time course of motion, the movement route a neuron _t-1 and the first time of t-1 Related neuron a movement route _t to t-th time (e.g., current motion path) related to each neuron in accordance with the generation time pieces movement route (e.g., movement route extracting module 140 extracts from the trajectory set movement route Or the time at which the exercise route extraction module 140 stores the extracted exercise route in the data center 130). In the LSTM network, the fusion of the intersection points of the motion routes is gradually carried out over time.

6, each neuron comprises input and output, to neurons and neurons A _t A _t-1 as an example, X _t denotes an input neuron A _t, which includes information related to the neuronal A _t motion path (e.g., the current movement path) of the intersection set, h _t represents the output neuron a _t, which comprises determining by neurons a _t optimized movement path set intersection, X _t-1 represents neurons a _t-1 The input of neuron A _t-1 , which includes the set of intersection points of the movement route related to neuron A t-1 (for example, a historical movement route before the current movement route), h _t-1 represents the output of neuron A _{t-1, which includes} The set of intersection points of the optimized motion route determined by the neuron At _-1. In addition, the neuron at the previous moment can transmit its cell state and output to the neuron at the next moment. Among them, the cell state is the memory of the LSTM network, and the cell state transmitted to a neuron can include the state and output of the neuron before the neuron. A collection of intersection points of multiple motor routes related to a neuron, and the cell state transmitted from one neuron to the next neuron may include the intersection of multiple motor routes related to the neuron and the neurons before the neuron. set. For example, neurons A _t-1 can be transmitted to the neurons neurons A A _t _t-1 _t-1 h, the output state of the cell, and C _t-1. Therefore, the output and cell state of each neuron is not only related to the input of the neuron, but also related to the output and cell state of the previous neuron. For example, the output h _{t-1 of} _{neuron A t-1} is determined by The set of intersection points of the optimized motion route determined by the neuron At _-1 will include one or more intersection points of the historical motion route from the 1st moment to the t-1th moment. The output h _{t of the} _{neuron A t} is determined by the neuron movement route optimization set intersection membered _t a determination may include a history of the current course of motion and the movement route of the first time t 1 to time t-1 the first time or a plurality of intersections.

Each neuron output LSTM network and forgotten cell state controlled by the gate, an input gate and an output gate, shown in Figure 6, an example in neurons A _t, f _t forgotten door decided neurons A _t-1 Which intersection points are discarded in the cell state C _t-1 _{, the input gate i t} determines which intersection points are added to the cell state C _t-1 of the neuron A _t-1 _{, and the output gate o t} determines the cell state C from the neuron A _t which intersection _t output to obtain an output h _t. Among them, the calculation formula of the _{forget gate f t is as follows:}

f _t =σ(W _f ·[h _t-1 ,X _t ]+b _f ) Formula 5

b _f =1+b _m formula 6

Among them, σ represents the sigmoid function; W _f represents the weight matrix of the forgetting gate; [h _t-1 ,X _t ] represents the long vector formed by splicing _{h t-1} and X _t _{; b f} represents the offset value of the forgetting gate; b _m represents the abnormal segment deviation calculated when determining the model element of the current motion route. It should be noted that b _{m of} each neuron represents the abnormal segment deviation calculated when determining the model element of the historical motion route related to each neuron. .

Neurons formula A _t i _t input gate follows

i _t =σ(W _i ·[h _t-1 ,X _t ]+b _i ) Formula 7

Among them, σ represents the sigmoid function; W _i represents the weight matrix of the input gate; [h _t-1 ,X _t ] represents the long vector formed by splicing _{h t-1} and X _t _{; b i} represents the bias value of the input gate.

A _t neurons, the cell state can be obtained i _t C _t A _t neurons via forgetting input gate door and f _t, the specific calculation formula is as follows:

among them,

Represents the input cell state; W _c represents the weight matrix of the _{input cell state; b c} represents the bias value of the input cell state;

Represents the input gate i _t control

Which intersection points will be updated to C _t ; f _t ·C _t-1 indicates which intersection points in the cell state C _t-1 of the forgetting gate f _t control neuron A _t-1 will be discarded, the higher the value of bm indicates the current GPS movement route worse signal quality, then the neuronal cells a _t-1 state C _t-1 by forgetting gate f _t discarded intersection will be less.

A neuron output gate o _t _t is calculated as follows:

o _t =σ(W _o ·[h _t-1 ,X _t ]+b _o ) Formula 10

Wherein, W _o represents the weight matrix of _{the output gate; b o} represents the bias value of the output gate.

A _t the neurons, the output of the gate which may be the intersection point C _t o _t to be output via the control state of the cell to obtain an output h _t, specifically calculated as follows:

h _t =o _t ·tanh(C _t ) Formula 11

It should be noted that although the LSTM network is taken as an example in FIG. 6 to describe the fusion of the intersection of the current movement route and each historical movement route, the movement route optimization module 150 of FIG. 2 can use the GRU network and implement it based on similar principles. The fusion of the current movement route and the intersection of each historical movement route.

Returning to Figure 2, the model feature determination unit 144 may also be used in accordance with the movement route model to determine the movement route optimization model element A _t neuron output. In particular, the model feature determination unit 144 may determine the intersection of the link table link relationship between each intersection in the intersection set a movement route optimization neuron output in A _t least one other point of intersection, for example, but not limited to, such as Figure 4 shows the doubly linked list between the intersection points. Since the movement of the intersection of the route optimization module 150 of the current history of the movement route and at least one course of motion were fused, and therefore the relationship between the link pieces of the original motion path intersection is broken, the model determination unit 144 may feature A _t from neural element In the set of intersection points of the output optimized motion route, select an intersection point as the starting point of the optimized motion route, calculate the distance between the remaining intersection points and the starting point (for example, but not limited to the Euclidean distance between two points), and determine the distance from the starting point The nearest intersection is used as the second intersection of the optimized movement route, and then the intersection closest to the second intersection is determined from the remaining intersections as the third intersection of the optimized movement route, and so on, until the last intersection is determined To optimize the end point of the exercise route. Intersection link movement route optimization table A _t the output neurons, each of the intersections other than the start and end points may include GPS coordinates of the intersection, the intersection of a front link (e.g., but not limited to, a pointer) and The link to the next point of intersection (for example, but not limited to, pointer), and the starting point may include the GPS coordinates of the starting point, the link to the next point of intersection (for example, but not limited to, pointer), and the end point may include the GPS coordinates of the end point, to The link to the previous intersection (for example, but not limited to, pointer).

Model element determination unit 144 may also determine which value is characterized by the optimized coding geohash movement route A _t neuron output raster plurality of geographical range. In particular, the model feature determination unit 144 may determine the minimum bounding rectangle of the movement route optimization output neuron A _t, wherein, the smallest circumscribed rectangle in which two parallel sides and a set of weft intersections movement route optimization most South The latitude coordinates of the northernmost point of intersection are determined, and the other two sides are parallel to the meridian and are determined by the longitude coordinates of the easternmost and westernmost intersection points in the set of intersection points of the optimized motion route. The model element determination unit 144 can determine the minimum circumscribed rectangle of the optimized motion route. center as the movement route optimization by the output neuron a _t the center position.

Model determination unit 144 may be elements of minimum bounding rectangle center position of the movement route optimization neuron output A _t the center, the movement route optimization divided into four grid, each grid is then determined corresponding to the geographical range geohash encoding value.

Model element determination unit 144 may also be in a central location of neurons A _t outputted movement route optimization origin construct relative coordinate system, by the absolute GPS coordinates of each intersection optimized movement route neuron A _t output (e.g., longitude And latitude) are converted to relative coordinates in the relative coordinate system.

Model element determination unit 144 may also be determined according to the motion path model, each element of the movement route optimization A _t neuron output stored in the data center 130, as the range of the current geographic movement route and the pieces of the history of the movement route which The default exercise route for and the candidate exercise route to be recommended to the user.

According to some embodiments of the present application, the sports route recommendation module 160 may receive a user's recommendation request for a sports route, obtain the user's current GPS coordinates, and calculate the geohash code value of the user's GPS coordinates. Then, the sports route recommendation module 160 can determine the grid geohash code value with the smallest difference from the geohash code value of the user’s GPS coordinates among the multiple grid geohash code values of the multiple sports routes stored in the data center 130. The geohash code value can represent a similar geographic location. Therefore, the grid geohash code value corresponding to the grid geohash code value with the smallest gap with the user’s GPS coordinate geohash code value is closest to the user’s geographic location, and the exercise route recommendation module 160 can recommend to the user The movement route. In an example, when the coordinates of the intersection of the recommended exercise route are GPS coordinates, the exercise route recommendation module 160 may present the recommended exercise route on the map according to the intersection link table of the recommended exercise route; at the intersection of the recommended exercise route In the case where the coordinates of are relative to the center position, the exercise route recommendation module 160 can present the recommended exercise route on the map according to the GPS coordinates of the center position and the intersection link table of the recommended exercise route.

In the embodiment of the present application, by segmenting, clustering, and extracting multiple motion trajectories generated by the user over a period of time, and extracting a simplified motion route, the storage and processing cost of the motion trajectory can be reduced, and the motion can be improved. The processing efficiency of the trajectory.

Further, by dividing the geographic area containing the movement route into multiple grids, and the multiple grids share a central position, the storage and processing cost of the movement track can be further reduced, and the geohash encoding of multiple grids can be used. The value index of the sports route can improve the retrieval efficiency of the sports route.

Further, by taking the center position of the movement route as the origin, and performing relative coordinate conversion on each intersection of the movement route, the privacy of the user can be protected and the sensitive data related to the user can be eliminated.

Further, when fusing the intersection of the current movement route and the historical movement route through the cyclic neural network, the abnormal segment deviation of the current movement route is considered, that is, the quality of the GPS signal when recording the movement trajectory related to the current movement route is considered. When the GPS signal quality is poor, the discarded historical movement routes will have fewer intersections, so the fusion error can be controlled more accurately, and the quality of the optimized movement route can also be improved.

Further, in the embodiment of the present application, the terminal device 100 is a user-side device. Compared with the processing of the movement route on the cloud-side server in the prior art, the embodiment of the present application can reduce the inability to cause network and cloud-side failures. Deal with the risk of the sports route, and can reduce the response delay to the user's route recommendation request.

Fig. 7 shows a schematic flowchart of a method for extracting a movement route according to an embodiment of the present invention. The motion route extraction module 140 of the terminal device 100 in FIG. 2 may implement different blocks or other parts of the method. For the content not described in the foregoing device embodiment, refer to the following method embodiment, and similarly, for the content not described in the method embodiment, refer to the foregoing device embodiment. As shown in Fig. 7, the method for extracting the movement route may include:

S701. Through the track segmentation unit 141, obtain from the data center 130 that the user recorded by the motion track recording module 120 is in the same geographic area within a period of time (for example, one day, three days, one week, two weeks, one month, or other time period) Multiple trajectories generated within (for example, the same sports field).

Examples of multiple motion trajectories may be, but are not limited to, the motion trajectories generated by the user today, yesterday, and the day before, or multiple motion trajectories obtained by dividing the motion trajectory generated by the user in one exercise, for example, the user The motion trajectory generated by running multiple laps along the sports field is divided into multiple motion trajectories according to each lap.

S702: Using the trajectory dividing unit 141, each of the multiple movement trajectories is divided into multiple trajectory segments.

Specifically, each trajectory can be expressed as

Represents the i-th trajectory point of the m-th motion trajectory,

S703: Through the trajectory clustering unit 142, the trajectory segments of each motion trajectory are sequentially clustered into a plurality of trajectory segment clusters, wherein, within a trajectory segment cluster, the distance between adjacent trajectory segments is less than a predetermined value, wherein In a cluster of trajectory segments, the distance between adjacent trajectory segments is less than a predetermined value, wherein the adjacent trajectory segments respectively belong to different running trajectories of the multiple running trajectories.

As an example, two trajectory segments are represented as ts _i and ts _j , where the starting point of the trajectory segment ts _i _{is si} , the end point is e _i , the starting point of the trajectory segment ts _j _{is s j} , and the end point is e _j , then The distance between _{trajectory segments ts i} and ts _j can be calculated based on the following formula:

d _|| = MIN(l _||1 , l _||2 ) Formula 14

By analogy, the trajectory clustering unit 142 of FIG. 2 can determine other trajectory segment clusters of the motion trajectory. It should be noted that, for the motion trajectory that is different from that shown in FIG. 3, each trajectory segment cluster of the motion trajectory can be determined based on the principle similar to the foregoing.

S704: For each trajectory segment cluster, the trajectory extraction unit 143 calculates a cluster vector based on each trajectory segment in the cluster.

As an example, the trajectory extraction unit 143 may use the average value of the GPS coordinates of the starting trajectory point of each trajectory segment in the trajectory segment cluster as the GPS coordinates of the starting point of the cluster vector, and the GPS coordinates of the ending trajectory point of each trajectory segment The mean value is used as the GPS coordinates of the end point of the cluster vector. As another example, the trajectory extraction unit 143 may use any trajectory segment in the trajectory segment cluster as the cluster vector. It should be noted that the cluster vector can also be determined based on any other suitable method.

S705: Through the trajectory clustering unit 142, determine the abnormal trajectory segments in the multiple motion trajectories, and determine the proportion of the abnormal trajectory segments in all the trajectory segments, that is, the abnormal segment deviation.

Specifically, the trajectory clustering unit 142 may use a hierarchical clustering algorithm, a K-means clustering algorithm or other clustering algorithms to cluster the center points of each trajectory segment of the multiple running trajectories, and obtain multiple trajectory segment center points. The trajectory clustering unit 142 can calculate the distance between the center point of each trajectory segment of multiple running trajectories and the center of each category, and the minimum distance between the center point and the center of each category is greater than a predetermined A trajectory segment with a value (the predetermined value is the same as or different from the aforementioned predetermined value related to the distance between trajectory segments) is marked as an abnormal trajectory segment. For example, in the example of FIG. 3, the trajectory segment P15 of _{TR 1} _{and the trajectory segment P35 of TR 3} may be marked as abnormal trajectory segments. In some cases, a higher abnormal segment deviation can indicate poor GPS signal quality during movement.

In S706, the model element determining unit 144 determines the intersection of the adjacent cluster vectors of the adjacent trajectory segment clusters to determine the set of intersections of the simplified motion routes extracted from the motion trajectory.

As an example, in the case where the adjacent cluster vectors of adjacent trajectory segment clusters do not meet end to end, the model element determining unit 144 may extend one or two of the adjacent cluster vectors to determine the adjacent cluster vector The GPS coordinates of the intersection between the adjacent cluster vectors; in the case where the adjacent cluster vectors have intersected, the model element determination unit 144 can directly determine the GPS coordinates of the intersection between the adjacent cluster vectors.

In S707, the model element determining unit 144 determines the start point and the end point of the motion route in the set of intersections of the motion route.

As an example, the model element determining unit 144 may use the starting point of the cluster vector of the first trajectory segment cluster of the motion trajectory as the starting point of the motion route, and the end point of the cluster vector of the last trajectory segment cluster as the end point of the motion route.

S708: Determine, through the model element determining unit 144, an intersection link table indicating a link relationship between each intersection in the set of intersections of the motion route and at least one other intersection, where the set of intersections includes the start and end points of the motion route.

As an example, the model element determining unit 144 may determine the doubly linked list between the intersection points as shown in FIG. 4, where the starting point may be the starting point of the motion route, and the intersection point 1 may be the cluster of the first trajectory segment cluster of the motion trajectory. The intersection point between the vector and the cluster vector of the second trajectory segment cluster. Intersection point 2 can be the intersection point between the cluster vector of the second trajectory segment cluster of the motion trajectory and the cluster vector of the third trajectory segment cluster, and so on, The end point can be the end point of the movement route. In addition, each intersection shown in FIG. 4 may include the GPS coordinates of the intersection, a link to the previous intersection (for example, but not limited to, a pointer), and a link to the next intersection (for example, but not limited to, a pointer) , And the starting point may include the GPS coordinates of the starting point and a link to the next point of intersection (for example, but not limited to, a pointer), and the end point may include the GPS coordinates of the ending point, and a link to the previous point of intersection (for example, but not limited to, a pointer). The doubly linked list shown in FIG. 4 enables the user to move along the forward direction of the movement route (for example, from the starting point shown in FIG. 4 to the end point shown in FIG. 4), or along the reverse direction of the movement route (for example, from the figure The end point shown in 4 points to the starting point shown in Fig. 4) to move. It should be noted that the model element determining unit 144 may also determine a one-way linked list between intersections, that is, each intersection shown in FIG. 4 may include the GPS coordinates of the intersection and the link to the previous intersection (for example, but not Limited to pointers) or links to the next intersection (for example, but not limited to pointers).

S709: Determine the center position of the motion route through the model element determination unit 144.

In an example, the model element determination unit 144 may determine the minimum circumscribed rectangle of the movement route, where two sides of the minimum circumscribed rectangle are parallel to the latitude and are determined by the latitude coordinates of the southernmost and northernmost intersection points in the set of intersection points of the movement route. It is determined that the other two sides are parallel to the meridian and are determined by the longitude coordinates of the easternmost and westernmost intersection points in the set of intersection points of the movement route, and the model element determination unit 144 may use the center point of the smallest bounding rectangle of the movement route as the center of the movement route. The position, for example, the GPS coordinates of the center position of the movement route shown in FIG. 5 are (121.623, 29.855).

In S710, the model element determining unit 144 determines geohash coding values of multiple grids representing the geographic range of the movement route.

In an example, as shown in FIG. 5, the model element determining unit 144 of FIG. 2 may take the center position as the center, divide the minimum bounding rectangle of the movement route into four grids, and determine the geographic location of each grid. The geohash code value corresponding to the range, where the number of bits of the geohash code value can be, for example, but not limited to, 7-bit base32 coding. For example, as shown in FIG. 5, the coding values of the four grids are wx4g340, wx4g341, wx4g342, and wx4g343.

S711: The model element determining unit 144 determines the relative coordinates of each intersection of the motion route with respect to the center position.

As an example, the model element determination unit 144 may use the aforementioned center position as the origin to construct a relative coordinate system, and convert the absolute GPS coordinates (for example, longitude and latitude) of each intersection of the movement route into relative coordinates in the relative coordinate system. Coordinates, in this way, can eliminate sensitive data related to users.

S712, through the model element determining unit 144, save each element of the simplified motion route extracted from the motion trajectory determined according to the motion route model in the data center 130 as a candidate motion route to be recommended to the user.

Fig. 8 shows a schematic flowchart of a method for optimizing a movement route according to an embodiment of the present invention. The motion route optimization module 150 and the motion route extraction module 140 of the terminal device 100 in FIG. 2 may implement different blocks or other parts of the method. For the content not described in the foregoing device embodiment, refer to the following method embodiment, and similarly, for the content not described in the method embodiment, refer to the foregoing device embodiment. As shown in Figure 8, the optimization method of the movement route may include:

S801, through the sports route optimization module 150, obtain a current sports route and at least one historical sports route located in the same geographic area (for example, the same sports field).

Among them, the current exercise route refers to the current route generated by the user in the current time period (for example, but not limited to, one day, three days, one week, two weeks, one month, or other time periods) by the method shown in FIG. The exercise route extracted from the exercise track set, each historical exercise route in at least one historical exercise route refers to the historical time period from the user before the current time period (for example, but not limited to, One day, three days, one week, two weeks, one month, or other time period) generated in the historical movement trajectory set extracted from the movement route.

S802, through the exercise route optimization module 150, input the intersection set of the current exercise route and the intersection set of each historical exercise route in the at least one historical exercise route into different neurons of the cyclic neural network to obtain the intersection set of the optimized exercise route . Among them, the recurrent neural network can refer to the above description related to FIG. 6, which will not be repeated here.

In S803, the model element determining unit 144 determines the intersection link table of the optimized motion route output by the last neuron of the recurrent neural network.

In the example of Figure 6, the unit 144 is determined by the model elements, model according to the movement route, the movement route optimization is determined neuron output A _t intersection link table. In particular, the model feature determination unit 144 may determine the intersection of the link table link relationship between each intersection in the intersection set a movement route optimization neuron output in A _t least one other point of intersection, for example, but not limited to, such as Figure 4 shows the doubly linked list between the intersection points. Since the movement of the intersection of the route optimization module 150 of the current history of the movement route and at least one course of motion were fused, and therefore the relationship between the link pieces of the original motion path intersection is broken, the model determination unit 144 may feature A _t from neural element In the set of intersection points of the output optimized motion route, select an intersection point as the starting point of the optimized motion route, calculate the distance between the remaining intersection points and the starting point (for example, but not limited to the Euclidean distance between two points), and determine the distance from the starting point The nearest intersection is used as the second intersection of the optimized movement route, and then the intersection closest to the second intersection is determined from the remaining intersections as the third intersection of the optimized movement route, and so on, until the last intersection is determined To optimize the end point of the exercise route. Intersection link movement route optimization table A _t the output neurons, each of the intersections other than the start and end points may include GPS coordinates of the intersection, the intersection of a front link (e.g., but not limited to, a pointer) and The link to the next point of intersection (for example, but not limited to, pointer), and the starting point may include the GPS coordinates of the starting point, the link to the next point of intersection (for example, but not limited to, pointer), and the end point may include the GPS coordinates of the end point, to The link to the previous intersection (for example, but not limited to, pointer).

S804: Optionally, the model element determining unit 144 determines the geohash encoding values of multiple grids representing the geographic range of the optimized motion route output by the last neuron of the recurrent neural network.

In the example of Figure 6, the elements of the model determination unit 144 may determine the minimum bounding rectangle of the movement route optimization output neuron A _t, wherein two of the smallest circumscribed rectangle sides parallel to weft and the movement route optimization by the intersection of a set of The latitude coordinates of the southernmost and northernmost intersection points in the middle are determined, and the other two sides are parallel to the meridian and are determined by the longitude coordinates of the easternmost and westernmost intersection points in the set of intersection points of the optimized motion route. The model element determination unit 144 can determine the optimal motion route's latitude coordinates. the center point of the smallest circumscribed rectangle as the center position of the movement route optimization a _t neuron output.

In S805, optionally, the model element determining unit 144 determines the relative coordinates of each intersection of the optimized motion route output by the last neuron of the cyclic neural network.

In the example of Figure 6, the determination unit 144 by the model elements, it may be the center position of neurons A _t outputted movement route optimization origin construct relative coordinate system, the respective intersections optimized movement route by neurons A _t output The absolute GPS coordinates (for example, longitude and latitude) of is converted to relative coordinates in the relative coordinate system.

In S806, the model element determining unit 144 stores the model element of the optimized motion route output by the last neuron of the recurrent neural network.

In the example of Figure 6, the unit 144 is determined by the model elements can be stored by the respective elements of the model movement route optimization A _t the output neuron in the data center 130, and the movement route as the current history of the course of motion in which the pieces of geographical range The default exercise route within.

Fig. 9 shows a schematic flowchart of a method for recommending a sports route according to an embodiment of the present invention. The motion route recommendation module 160 of the terminal device 100 in FIG. 2 may implement different blocks or other parts of the method. For the content not described in the foregoing device embodiment, refer to the following method embodiment, and similarly, for the content not described in the method embodiment, refer to the foregoing device embodiment. As shown in Figure 7, the recommended method of exercise route may include:

S901: Through the sports route recommendation module 160, a user's recommendation request for a sports route is received.

S902: Obtain the current GPS coordinates of the user from the GPS positioning module 120 through the exercise route recommendation module 160.

S903: Calculate the geohash code value of the GPS coordinates of the user through the exercise route recommendation module 160.

S904, through the exercise route recommendation module 160, among the multiple raster geohash code values of the multiple exercise routes stored in the data center 130, determine the raster geohash code value with the smallest difference with the geohash code value of the user's GPS coordinates;

S905, through the exercise route recommendation module 160, recommend to the user the exercise route corresponding to the grid geohash code value with the smallest difference between the geohash code value of the GPS coordinates of the user.

Since similar geohash code values can indicate similar geographic locations, the motion route corresponding to the grid geohash code value with the smallest gap between the geohash code value of the user’s GPS coordinates and the user’s geographic location is closest to the user’s geographic location, and the motion route recommendation module 160 can The user recommends the exercise route. In an example, when the coordinates of the intersection of the recommended exercise route are GPS coordinates, the exercise route recommendation module 160 may present the recommended exercise route on the map according to the intersection link table of the recommended exercise route; at the intersection of the recommended exercise route In the case where the coordinates of are relative to the center position, the exercise route recommendation module 160 can present the recommended exercise route on the map according to the GPS coordinates of the center position and the intersection link table of the recommended exercise route.

In the embodiment of the present application, by segmenting, clustering, and extracting the motion trajectory generated by the user over a period of time, and extracting a simplified motion route, the storage and processing cost of the motion trajectory can be reduced, and the performance of the motion trajectory can be improved. Processing efficiency.

FIG. 10 shows a schematic structural diagram of a movement route processing device 1000 according to an embodiment of the present application. The device 1000 may include one or more processors 1002, a system control logic 1008 connected to at least one of the processors 1002, a system memory 1004 connected to the system control logic 1008, and a non-volatile memory connected to the system control logic 1008 (NVM) 1006, and a network interface 1010 connected to the system control logic 1008.

The processor 1002 may include one or more single-core or multi-core processors. The processor 1002 may include any combination of a general-purpose processor and a special-purpose processor (for example, a graphics processor, an application processor, a baseband processor, etc.). In the embodiment of the present application, the processor 1002 may be configured to execute one or more embodiments according to the various embodiments shown in FIGS. 7-9.

In some embodiments, the system control logic 1008 may include any suitable interface controller to provide any suitable interface to at least one of the processors 1002 and/or any suitable device or component in communication with the system control logic 1008.

In some embodiments, the system control logic 1008 may include one or more memory controllers to provide an interface to the system memory 1004. The system memory 1004 can be used to load and store data and/or instructions. In some embodiments, the memory 1004 of the device 1000 may include any suitable volatile memory, such as a suitable dynamic random access memory (DRAM).

The NVM/memory 1006 may include one or more tangible, non-transitory computer-readable media for storing data and/or instructions. In some embodiments, the NVM/memory 1006 may include any suitable non-volatile memory such as flash memory and/or any suitable non-volatile storage device, such as HDD (Hard Disk Drive, hard disk drive), CD (Compact Disc , At least one of an optical disc drive and a DVD (Digital Versatile Disc, Digital Versatile Disc) drive.

The NVM/memory 1006 may include a part of storage resources installed on the apparatus of the apparatus 1000, or it may be accessed by the device, but not necessarily a part of the device. For example, the NVM/storage 1006 can be accessed through the network via the network interface 1010.

In particular, the system memory 1004 and the NVM/memory 1006 may respectively include: a temporary copy and a permanent copy of the instruction 1020. The instructions 1020 may include instructions that, when executed by at least one of the processors 1002, cause the apparatus 1000 to implement the method shown in FIGS. 7-9. In some embodiments, instructions 1020, hardware, firmware, and/or software components thereof may additionally/alternatively be placed in system control logic 1008, network interface 1010, and/or processor 1002.

The network interface 1010 may include a transceiver to provide a radio interface for the device 1000, and then communicate with any other suitable equipment (such as a front-end module, an antenna, etc.) through one or more networks. In some embodiments, the network interface 1010 may be integrated with other components of the device 1000. For example, the network interface 1010 may be integrated in at least one of the processor 1002, the system memory 1004, the NVM/storage 1006, and a firmware device with instructions (not shown), when at least one of the processor 1002 executes the When instructed, the device 1000 implements one or more of the various embodiments shown in FIGS. 7-9.

The network interface 1010 may further include any suitable hardware and/or firmware to provide a multiple input multiple output radio interface. For example, the network interface 1010 may be a network adapter, a wireless network adapter, a telephone modem and/or a wireless modem.

In one embodiment, at least one of the processors 1002 may be packaged with the logic of one or more controllers used for the system control logic 1008 to form a system in package (SiP). In one embodiment, at least one of the processors 1002 may be integrated with the logic of one or more controllers used for the system control logic 1008 on the same die to form a system on chip (SoC).

The device 1000 may further include: an input/output (I/O) interface 1012. The I/O interface 1012 may include a user interface to enable the user to interact with the device 1000; the design of the peripheral component interface enables the peripheral components to also interact with the device 1000. In some embodiments, the device 1000 further includes a sensor for determining at least one of environmental conditions and location information related to the device 1000.

In some embodiments, the user interface may include, but is not limited to, a display (e.g., liquid crystal display, touch screen display, etc.), speakers, microphones, one or more cameras (e.g., still image cameras and/or video cameras), flashlights (e.g., LED flash) and keyboard.

In some embodiments, the peripheral component interface may include, but is not limited to, a non-volatile memory port, an audio jack, and a power interface.

In some embodiments, the sensor may include, but is not limited to, a gyroscope sensor, an accelerometer, a proximity sensor, an ambient light sensor, and a positioning unit. The positioning unit may also be part of or interact with the network interface 1010 to communicate with components of the positioning network (eg, global positioning system (GPS) satellites).

Although the description of this application will be introduced in conjunction with the preferred embodiments, this does not mean that the features of the invention are limited to this embodiment. On the contrary, the purpose of introducing the invention in combination with the embodiments is to cover other options or modifications that may be extended based on the claims of this application. In order to provide an in-depth understanding of the application, the following description will contain many specific details. This application can also be implemented without using these details. In addition, in order to avoid confusion or obscuring the focus of this application, some specific details will be omitted in the description. It should be noted that the embodiments in the application and the features in the embodiments can be combined with each other if there is no conflict.

In addition, various operations will be described as a plurality of discrete operations in a manner that is most helpful for understanding the illustrative embodiments; however, the order of description should not be construed as implying that these operations must depend on the order. In particular, these operations need not be performed in the order of presentation.

Unless the context dictates otherwise, the terms "comprising", "having" and "including" are synonymous. The phrase "A/B" means "A or B". The phrase "A and/or B" means "(A and B) or (A or B)".

As used herein, the term "module" or "unit" can refer to, be, or include: application specific integrated circuit (ASIC), electronic circuit, (shared, dedicated, or group) processing that executes one or more software or firmware programs And/or memory, combinatorial logic circuits, and/or other suitable components that provide the described functions.

In the drawings, some structural or method features are shown in a specific arrangement and/or order. However, it should be understood that such a specific arrangement and/or ordering may not be required. In some embodiments, these features may be arranged in a different manner and/or order than shown in the illustrative drawings. In addition, the inclusion of structural or method features in a particular figure does not imply that such features are required in all embodiments, and in some embodiments, these features may not be included or may be combined with other features.

The various embodiments of the mechanism disclosed in this application may be implemented in hardware, software, firmware, or a combination of these implementation methods. The embodiments of the present application can be implemented as a computer program or program code executed on a programmable system including at least one processor and a storage system (including volatile and non-volatile memory and/or storage elements) , At least one input device and at least one output device.

Program codes can be applied to input instructions to perform the functions described in this application and generate output information. The output information can be applied to one or more output devices in a known manner. For the purposes of this application, a processing system includes any system having a processor such as, for example, a digital signal processor (DSP), a microcontroller, an application specific integrated circuit (ASIC), or a microprocessor.

The program code can be implemented in a high-level programming language or an object-oriented programming language to communicate with the processing system. When needed, assembly language or machine language can also be used to implement the program code. In fact, the mechanism described in this application is not limited to the scope of any particular programming language. In either case, the language can be a compiled language or an interpreted language.

In some cases, the disclosed embodiments may be implemented in hardware, firmware, software, or any combination thereof. In some cases, one or more aspects of at least some embodiments may be implemented by representative instructions stored on a computer-readable storage medium. The instructions represent various logics in the processor, and the instructions, when read by a machine, cause This machine makes the logic used to execute the techniques described in this application. These representations called "IP cores" can be stored on a tangible computer-readable storage medium and provided to multiple customers or production facilities to be loaded into the manufacturing machine that actually manufactures the logic or processor.

Such computer-readable storage media may include, but are not limited to, non-transitory tangible arrangements of objects manufactured or formed by machines or equipment, including storage media, such as hard disks, any other types of disks, including floppy disks, optical disks, compact disks, etc. Disk read only memory (CD-ROM), compact disk rewritable (CD-RW), and magneto-optical disk; semiconductor devices such as read only memory (ROM), such as dynamic random access memory (DRAM) and static random access Random access memory (RAM) such as memory (SRAM), erasable programmable read-only memory (EPROM), flash memory, electrically erasable programmable read-only memory (EEPROM); phase change memory (PCM); magnetic card Or optical card; or any other type of medium suitable for storing electronic instructions.

Therefore, each embodiment of the present application also includes a non-transitory computer-readable storage medium, which contains instructions or contains design data, such as hardware description language (HDL), which defines the structures, circuits, devices, etc. described in the present application. Processor and/or system characteristics.

Claims

A method for processing motion trajectory, which is characterized in that it comprises:

Acquiring a set of motion trajectories, wherein the set of motion trajectories includes a plurality of motion trajectories; and

Determining a movement route element according to the movement trajectory set, wherein the movement route element includes at least one of a center position and an intersection set,

Wherein, the center position is shared by multiple grids and the set of motion trajectories is contained within the geographic range of the multiple grids; and

Wherein, the set of intersection points includes a plurality of intersection points related to a plurality of trajectory segment clusters, wherein each trajectory segment cluster in the plurality of trajectory segment clusters includes a plurality of trajectory segments of the multiple motion trajectories, and the The distance between adjacent trajectory segments in the plurality of trajectory segments is less than a predetermined value, wherein the adjacent trajectory segments respectively belong to different trajectories of the multiple trajectories.
8. The motion track processing method of claim 1, wherein the motion route element further comprises:

The Geohash code corresponding to each grid of the plurality of grids is used to index the geographic range corresponding to each grid.
The motion track processing method according to claim 2, wherein the Geohash code includes at least 7-bit Base32 coded characters.
The motion trajectory processing method according to claim 1, wherein the motion route element further comprises: indicating the distance between each intersection of the plurality of intersections and at least one other intersection of the plurality of intersections. Intersection link table of link relationship.
The motion track processing method according to claim 1, wherein the center position includes the GPS coordinates of the center position, and each of the multiple intersection points includes an origin with the center position as the origin. Relative coordinates in the relative coordinate system.
The method for processing motion trajectory according to claim 1, wherein the determining the motion route element according to the motion trajectory set comprises:

Selecting the four most east, south, west, and northest intersections from the multiple intersections;

The geographic range of the multiple grids is determined according to the four intersection points, and the center point of the geographic range is determined as the center position.
The method for processing motion trajectory according to claim 1, wherein the determining the motion route element according to the motion trajectory set comprises:

Dividing each motion trajectory in the motion trajectory set into multiple trajectory segments;

Clustering the plurality of trajectory segments whose distance between the adjacent trajectory segments is less than the predetermined value into one trajectory segment cluster among the plurality of trajectory segment clusters;

According to the multiple trajectory segments in each trajectory segment cluster in the multiple trajectory clusters, a cluster vector corresponding to each trajectory segment cluster is determined to form a cluster vector corresponding to the multiple trajectory segment clusters. Multiple cluster vectors;

Determine the intersection point between adjacent cluster vectors in the plurality of cluster vectors.
8. The motion trajectory processing method of claim 7, wherein the starting point coordinates of the cluster vector are the average value of the starting point coordinates of the multiple trajectory segments in the trajectory segment cluster, and the end point of the cluster vector The coordinate is the average value of the end point coordinates of the multiple trajectory segments in the trajectory segment cluster.
A method for processing motion trajectory, which is characterized in that it comprises:

Acquiring multiple current motion trajectory intersections related to the current motion trajectory set and multiple historical motion trajectory intersections related to the historical motion trajectory set; and

Determine whether to retain at least one current motion trajectory intersection of the multiple current motion trajectory intersections and/or at least one historical motion trajectory intersection of the multiple historical motion trajectory intersections at least partially based on the abnormal segment deviation, so as to obtain optimization Route of motion;

Wherein, the abnormal segment deviation at least partly specifies the proportion of abnormal trajectory segments in the multiple trajectory segments of the multiple trajectories in the current trajectory set, wherein the distance between the abnormal trajectory segment and the multiple trajectory segments at least A class center exceeds the first predetermined value.
9. The motion trajectory processing method according to claim 9, wherein the current motion trajectory set and the historical motion trajectory set are contained in the same geographic range.
9. The motion trajectory processing method according to claim 9, wherein the determining whether to retain at least one current motion trajectory intersection point among the multiple current motion trajectory intersection points and/or one of the multiple historical motion trajectory intersection points At least one historical motion trajectory intersection point, including using an LSTM (Long short-term memory) model to determine whether to retain at least one current motion trajectory intersection point and/or the multiple historical motions among the multiple current motion trajectory intersection points At least one historical movement trajectory intersection in the trajectory intersection.
The motion trajectory processing method according to any one of claims 9 to 11, wherein the determining whether to retain at least one current motion trajectory intersection point and/or the multiple current motion trajectory intersection points At least one historical motion trajectory intersection point among the historical motion trajectory intersection points further includes: using an LSTM model to calculate a forgetting gate for determining whether to retain the at least one historical motion trajectory intersection point among the multiple historical motion trajectory intersection points according to the following formula :

f t =σ(W f ·[h t-1 ,X t ]+b f )

b f =1+b m

Wherein, f t represents the forgetting gate of the neuron at the t-th moment of the LSTM model, σ represents the sigmoid function, W f represents the weight matrix of the forgetting gate, and X t represents the t-th neuron of the LSTM model The input of the neuron at time and includes the intersection of the multiple current motion trajectories, h t-1 represents the output of the neuron at the t-1 time of the LSTM model and includes the intersection of the multiple historical motion trajectories, [ h t-1 , X t ] represents a long vector formed by splicing h t-1 and X t , b f represents the offset value of the forgetting gate, and b m represents the deviation of the abnormal segment.
The motion trajectory processing method according to any one of claims 9 to 12, wherein the determining whether to retain at least one current motion trajectory intersection point and/or the multiple current motion trajectory intersection points At least one historical motion trajectory intersection point in the historical motion trajectory intersection points further includes using an LSTM model to calculate an input gate for determining whether to retain at least one current motion trajectory intersection point among the multiple current motion trajectory intersection points according to the following formula:

i t =σ(W i ·[h t-1 ,X t ]+b i )

Wherein, i t t represents the time of the model of the input gate LSTM neuron, σ represents a sigmoid function, W i represents the weight of the weight matrix input gate, X t denotes the t-th model LSTM The input of the neuron at time and includes the intersection of the multiple current motion trajectories, h t-1 represents the output of the neuron at the t-1 time of the LSTM model and includes the intersection of the multiple historical motion trajectories, [ h t-1 , X t ] represents a long vector formed by splicing h t-1 and X t , and b i represents the bias value of the input gate.
The motion trajectory processing method according to any one of claims 9 to 13, wherein the determining whether to retain at least one current motion trajectory intersection point and/or the multiple current motion trajectory intersection points The at least one historical motion trajectory intersection point in the historical motion trajectory intersection further includes: using an LSTM model to determine whether to output at least one of the multiple current motion trajectory intersection points and the multiple historical motion trajectory intersection points according to the following formula At least one of the historical motion trajectory intersection points in, so as to obtain the optimized motion route:

o t =σ(W o ·[h t-1 ,X t ]+b o )

h t ＝o t ·tanh(C t )

Wherein, X t represents the input of the neuron at the t-th time of the LSTM model and includes the intersection points of the multiple current motion trajectories, and h t-1 represents the input of the neuron at the t-1-th time of the LSTM model Output and include the intersection of the multiple historical motion trajectories,
Represents the input cell state of the neuron at the t-th time of the LSTM model, W c represents the weight matrix of the input cell state of the neuron at the t-th time of the LSTM model, and b c represents the weight matrix of the neuron at the t-th time of the LSTM model. The bias value of the input cell state of the neuron at time t, C t represents the cell state of the neuron at the t time of the LSTM model, and C t-1 represents the t-1 time of the LSTM model the cell state neurons, f t t represents the time of the neuron model LSTM forgotten door, i t t represents the time of the neuron model LSTM input gate, o t represents the The output gate of the neuron at the t-th time of the LSTM model, W o represents the weight matrix of the output gate, b o represents the bias value of the output gate, and h t represents the t-th time of the LSTM model The output of the neuron may also include multiple intersections of the optimized movement route.
The motion trajectory processing method according to claim 14, wherein the method further comprises selecting the most east, south, west, and north from the multiple intersection points of the optimized motion route Four intersections

According to the four intersection points, the geographic range including the movement route is determined, and the center point of the geographic range is determined as the center position of the optimized movement route.
The motion trajectory processing method according to claim 15, wherein the method further comprises determining the multiple grids of the geographic range and the geohash coding values of the multiple grids, wherein the multiple grids The grid shares the center position.
The motion track processing method of claim 15, wherein the center position includes the GPS coordinates of the center position, and each of the multiple intersection points of the optimized motion route includes the The center position is the relative coordinate in the relative coordinate system of the origin.
9. The motion trajectory processing method of claim 9, wherein the method further comprises:

Clustering the centers of the plurality of trajectory segments and obtaining the at least one cluster center of the plurality of trajectory segments.
A method for recommending sports routes, which is characterized in that it includes:

Receive user's recommendation request for sports route;

Acquiring the GPS coordinates of the user, and calculating the geohash code value of the GPS (Global Positioning System, global positioning system) coordinates of the user;

From the multiple grid geohash code values, query the grid geohash code value with the smallest difference with the geohash code value of the GPS coordinates of the user;

Obtain the movement route corresponding to the grid geohash code value with the smallest difference between the geohash code value of the GPS coordinates of the user and the center position of the movement route, wherein the center position is determined by the geographic location that contains the movement route. Multiple grid sharing of the range; and

Recommend the exercise route to the user according to the center position.
A computer-readable storage medium, characterized in that instructions are stored on the computer-readable storage, and when the instructions are executed on the machine, the machine executes any one of claims 1 to 19 The method described.
A motion trajectory processing device, characterized in that it comprises:

processor;

A memory, where instructions are stored on the memory, and when the instructions are executed by the processor, the processor is caused to execute the method according to any one of claims 1 to 19.
A motion trajectory processing device, characterized in that it comprises:

The motion track recording module is used to obtain a motion track set, wherein the motion track set includes a plurality of motion tracks; and

The movement route extraction module is configured to determine movement route elements according to the movement trajectory set, wherein the movement route elements include at least one of a center position and an intersection set,

Wherein, the center position is shared by multiple grids and the set of motion trajectories is contained within the geographic range of the multiple grids; and

Wherein, the set of intersection points includes a plurality of intersection points related to a plurality of trajectory segment clusters, wherein each trajectory segment cluster in the plurality of trajectory segment clusters includes a plurality of trajectory segments of the multiple motion trajectories, and the The distance between adjacent trajectory segments in the plurality of trajectory segments is less than a predetermined value, wherein the adjacent trajectory segments respectively belong to different trajectories of the multiple trajectories.
The motion trajectory processing device according to claim 22, wherein the motion route element further comprises:

The Geohash code corresponding to each grid of the plurality of grids is used to index the geographic range corresponding to each grid.
The motion track processing device of claim 23, wherein the Geohash code includes at least 7-bit Base32 coded characters.
The motion trajectory processing device according to claim 22, wherein the motion route element further comprises:

An intersection link table indicating a link relationship between each intersection of the plurality of intersections and at least one other intersection of the plurality of intersections.
The motion trajectory processing device according to claim 22, wherein the motion route extraction module is further configured to determine the relative coordinates of each of the multiple intersection points with the center position as the origin. Relative coordinates in the system.
The motion trajectory processing device according to claim 22, wherein the motion route extraction module is configured to determine the motion route element according to the motion trajectory set, including:

Selecting the four most east, south, west, and northest intersections from the multiple intersections;

The geographic range of the multiple grids is determined according to the four intersection points, and the center point of the geographic range is determined as the center position.
The motion trajectory processing device according to claim 22, wherein the motion route extraction module is configured to determine the motion route element according to the motion trajectory set, including:

Dividing each motion trajectory in the motion trajectory set into multiple trajectory segments;

Clustering the plurality of trajectory segments whose distance between the adjacent trajectory segments is less than the predetermined value into one trajectory segment cluster among the plurality of trajectory segment clusters;

According to the multiple trajectory segments in each trajectory segment cluster in the multiple trajectory clusters, a cluster vector corresponding to each trajectory segment cluster is determined to form a cluster vector corresponding to the multiple trajectory segment clusters. Multiple cluster vectors;

Determine the intersection point between adjacent cluster vectors in the plurality of cluster vectors.
The motion trajectory processing device according to claim 28, wherein the start point coordinates of the cluster vector are the average value of the start point coordinates of the multiple trajectory segments in the trajectory segment cluster, and the end point of the cluster vector The coordinate is the average value of the end point coordinates of the multiple trajectory segments in the trajectory segment cluster.
A motion trajectory processing device, characterized in that it comprises:

The motion route extraction module is used to obtain multiple current motion trajectory intersection points related to the current motion trajectory set and multiple historical motion trajectory intersection points related to the historical motion trajectory set; and

The motion route optimization module determines whether to retain at least one current motion trajectory intersection of the multiple current motion trajectory intersections and/or at least one historical motion trajectory of the multiple historical motion trajectory intersections at least partially based on the abnormal segment deviation Intersection point, so as to obtain an optimized movement route;

Wherein, the abnormal segment deviation at least partially specifies the proportion of abnormal trajectory segments in the multiple trajectory segments of the multiple motion trajectories in the current motion trajectory set, wherein the distance from the abnormal trajectory segment to at least A class center exceeds the first predetermined value.
The motion trajectory processing device according to claim 30, wherein the current motion trajectory set and the historical motion trajectory set are contained in the same geographic range.
The motion trajectory processing device according to claim 30, wherein the motion route optimization module is used to determine whether to retain at least one current motion trajectory intersection point and/or the multiple current motion trajectory intersection points. At least one historical motion trajectory intersection in the historical motion trajectory intersection includes a method for determining whether to retain at least one current motion trajectory intersection among the multiple current motion trajectory intersections by using an LSTM (Long short-term memory) model And/or at least one historical motion trajectory intersection among the multiple historical motion trajectory intersections.
The motion trajectory processing device according to any one of claims 30 to 32, wherein the motion route optimization module is used to determine whether to retain at least one current motion trajectory intersection and /Or at least one historical motion trajectory intersection point among the multiple historical motion trajectory intersection points, further comprising: using an LSTM model to calculate according to the following formula for determining whether to retain the at least one historical motion trajectory intersection A forgotten door at the intersection of historical motion trajectories:

f t =σ(W f ·[h t-1 ,X t ]+b f )

b f =1+b m

Wherein, f t represents the forgetting gate of the neuron at the t-th moment of the LSTM model, σ represents the sigmoid function, W f represents the weight matrix of the forgetting gate, and X t represents the t-th neuron of the LSTM model The input of the neuron at time and includes the intersection of the multiple current motion trajectories, h t-1 represents the output of the neuron at the t-1 time of the LSTM model and includes the intersection of the multiple historical motion trajectories, [ h t-1 , X t ] represents a long vector formed by splicing h t-1 and X t , b f represents the offset value of the forgetting gate, and b m represents the deviation of the abnormal segment.
The motion trajectory processing device according to any one of claims 30 to 33, wherein the motion route optimization module is used to determine whether to retain at least one current motion trajectory intersection and /Or at least one historical motion trajectory intersection among the multiple historical motion trajectory intersections, further comprising: using an LSTM model to calculate according to the following formula for determining whether to retain at least one current of the multiple current motion trajectory intersections The input gate at the intersection of the motion trajectory:

i t =σ(W i ·[h t-1 ,X t ]+b i )

Wherein, i t t represents the time of the model of the input gate LSTM neuron, σ represents a sigmoid function, W i represents the weight of the weight matrix input gate, X t denotes the t-th model LSTM The input of the neuron at time and includes the intersection of the multiple current motion trajectories, h t-1 represents the output of the neuron at the t-1 time of the LSTM model and includes the intersection of the multiple historical motion trajectories, [ h t-1 , X t ] represents a long vector formed by splicing h t-1 and X t , and b i represents the bias value of the input gate.
The motion trajectory processing device according to any one of claims 30 to 34, wherein the motion route optimization module is used to determine whether to retain at least one current motion trajectory intersection and /Or at least one historical motion trajectory intersection of the multiple historical motion trajectory intersections, further comprising: using an LSTM model to determine whether to output at least one current motion trajectory intersection of the multiple current motion trajectory intersections according to the following formula And at least one historical motion trajectory intersection point among the multiple historical motion trajectory intersection points, so as to obtain the optimized motion route:

o t =σ(W o ·[h t-1 ,X t ]+b o )

h t ＝o t ·tanh(C t )

Wherein, X t represents the input of the neuron at the t-th time of the LSTM model and includes the intersection points of the multiple current motion trajectories, and h t-1 represents the input of the neuron at the t-1-th time of the LSTM model Output and include the intersection of the multiple historical motion trajectories,
Represents the input cell state of the neuron at the t-th time of the LSTM model, W c represents the weight matrix of the input cell state of the neuron at the t-th time of the LSTM model, and b c represents the weight matrix of the neuron at the t-th time of the LSTM model. The bias value of the input cell state of the neuron at time t, C t represents the cell state of the neuron at the t time of the LSTM model, and C t-1 represents the t-1 time of the LSTM model the cell state neurons, f t t represents the time of the neuron model LSTM forgotten door, i t t represents the time of the neuron model LSTM input gate, o t represents the The output gate of the neuron at the t-th time of the LSTM model, W o represents the weight matrix of the output gate, b o represents the bias value of the output gate, and h t represents the t-th time of the LSTM model The output of the neuron may also include multiple intersections of the optimized movement route.
The motion trajectory processing device according to claim 35, wherein the motion route extraction module is further configured to select the easternmost, the southernmost, and the westernmost from the multiple intersection points of the optimized motion route. , The four northernmost intersections;

According to the four intersection points, the geographic range including the movement route is determined, and the center point of the geographic range is determined as the center position of the optimized movement route.
The motion trajectory processing device according to claim 36, wherein the motion route extraction module is further configured to determine multiple grids of the geographic range and geohash coding values of the multiple grids, wherein The multiple grids share the center position.
The motion trajectory processing device of claim 36, wherein the motion route extraction module is further configured to determine that each of the multiple intersection points of the optimized motion route is at the center position. It is the relative coordinate in the relative coordinate system of the origin.
The motion trajectory processing device of claim 30, wherein the motion route extraction module is further configured to:

Clustering the centers of the multiple trajectory segments and obtaining the at least one cluster center of the multiple trajectory segments.