An evolutionary clustering method and system for moving objects
Through the evolutionary clustering method for moving objects, the minimized group and density clustering algorithm and combined with the grid index technology, the problems of insufficient clustering quality and large time overhead in the existing technology are solved, and efficient and stable Clustering of moving objects, suitable for real-time scenarios.
Patent Information
- Application Number
- CN202411040058.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-31
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-07-31
AI Technical Summary
The existing real-time moving object clustering methods have shortcomings in ensuring clustering quality, especially the inability to effectively eliminate fluctuations in clustering, and the time overhead is too high when applied to moving objects, so they cannot adapt to the characteristics of fast moving.
Using an evolutionary clustering method for moving objects, the motion state information is obtained online, the minimization group is divided, the seed points are smoothed, and the parameters are optimized using density clustering algorithm to achieve clustering with approximate linear time complexity, combined with Grid indexing technology accelerates the search and clustering process.
It improves clustering quality and efficiency, can efficiently handle the clustering problem of moving objects in real-time scenarios, is suitable for real-time applications, reduces time overhead, and optimizes clustering parameters to improve the stability of the results.
Smart Images

Figure CN118568531B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of spatiotemporal trajectory data mining, and in particular relates to an evolutionary clustering method and system for moving objects. Background Art
[0002] With the advancement of technology, it is now easy for mobile objects, including people, vehicles, and ships, to transmit their locations to a central location in real time. This scenario opens up the possibility of discovering hidden movement patterns in real time. Clustering, a typical movement pattern discovery method, aims to group a group of moving objects based on their spatial-temporal similarities. The ability to continuously cluster moving objects is beneficial for a wide range of services and applications, such as ship collision risk assessment, path modeling, real-time road traffic management, and animal migration analysis.
[0003] Existing real-time clustering methods, such as [P. Kalnis, N. Mamoulis, and S. Bakiras, “Ondiscovering moving clusters in spatio-temporal data,” in SSTD. Springer, 2005, pp. 364–381], focus primarily on runtime efficiency and often neglect clustering quality, which can lead to unstable clustering results. The positions of moving objects typically change continuously and gradually, so clusters should evolve smoothly over time. Fluctuations in clustering are not only undesirable but can also have adverse effects on applications and services.
[0004] Existing real-time moving object clustering methods, such as [CS Jensen, D. Lin, and BC Ooi, “Continuous clustering of moving objects,” TKDE, vol. 19, no. 9, pp. 1161–1174, 2007], use the same clustering algorithm parameters to perform clustering across multiple time slices. However, in real-world scenarios, the optimal clustering algorithm parameters corresponding to different moments may be different. In addition, these methods cannot eliminate fluctuations in clustering. The key reason is that they perform clustering at the current timestamp without considering historical clustering. This problem can be solved by applying the concept of evolutionary clustering from the field of dynamic networks. By exploiting temporal smoothness, which can produce robustness to short-term variations, evolutionary clustering can achieve better clustering than traditional clustering techniques.
[0005] Existing evolutionary clustering studies, such as [DJ Fenn, MA Porter, M. McDonald, S.Williams, NF Johnson, and NS Jones, “Dynamic communities inmultichannel data: An application to the foreign exchange market during the2007–2008 credit crisis,” J Nonlinear Sci, vol. 19, no. 3, p. 033119, 2009], mainly target dynamic networks and are not suitable for moving objects. Using evolutionary clustering algorithms on moving objects is time-consuming for three main reasons: first, these solutions are designed specifically for dynamic networks, which are very different from spatiotemporal data; second, the movement speed of moving objects is usually much faster than the evolution speed of dynamic networks, which makes the temporal smoothness used in existing studies too strict for moving objects; third, existing studies usually perform clustering at each timestamp, which is computationally very expensive and infeasible for a large number of movements. Summary of the Invention
[0006] In view of the above, the present invention provides an evolutionary clustering method and system for moving objects, which can take into account the quality of clustering during the clustering process and introduce historical clustering results into clustering considerations, so that the evolutionary clustering algorithm can be executed with approximately linear time complexity on the moving object clustering problem, and the parameters of the density-based clustering algorithm can be optimized, and the optimized values of the parameters can be calculated in real time.
[0007] An evolutionary clustering method for moving objects includes the following steps:
[0008] (1) Obtain the motion state information of the moving object online. Each piece of motion state information includes the geographical coordinates (latitude and longitude) of the moving object and a timestamp;
[0009] (2) Determine the moving objects within the current time segment and divide these moving objects into multiple minimized groups;
[0010] (3) Based on the seed point of the minimization group, the positions of non-seed points in the group are smoothed;
[0011] (4) Cluster the moving objects in the current time segment according to the smoothed positions.
[0012] Furthermore, in step (2), when any piece of motion state information is obtained online, if the difference between the timestamp of the motion state information and the timestamp of the first piece of motion state information obtained in the current time segment is less than the set time interval, the moving object corresponding to the motion state information is classified as a moving object in the current time segment.
[0013] Furthermore, in step (2), for any moving object in the current time segment p , the geographical location coordinates in its motion state information are transformed into two-dimensional plane coordinates through the Miller projection method ( x , y ), find the p Distance less than md_thres All moving objects, if the number of these moving objects exceeds mn_thres , then p Form a minimum group for the initial seed point and divide all the moving objects into this group (and they cannot be snatched away by other seed point groups). md_thres is the distance threshold, mn_thres is the quantity threshold.
[0014] Furthermore, the specific implementation of step (3) is as follows:
[0015] (3.1) Any moving object in the group is used as a seed point, and the remaining moving objects are non-seed points. The seed points and non-seed points are speed-limited separately. Specifically:
[0016] For the seed point, by adjusting s _ loc Make d ( s _ loc,s _ loc ') / ( s _ t - s _ t ')< maxSpeed ,in s _ loc and s _ loc 'represents the position of the seed point in the current time segment and in the previous time segment, respectively. s _ t and s _ t 'Respectively s _ loc and s _ loc 'The corresponding timestamp, d ( s _ loc,s _loc ')express s _ loc and s _ loc 'The distance between maxSpeed is the maximum speed limit;
[0017] For any non-seed point o , then by adjusting o_loc So that it satisfies the following relationship:
[0018] ;
[0019] d ( sp,o _ loc ') / ( o _ t - o _ t ')< maxSpeed
[0020] in: o _ loc and o _ loc 'Represents non-seed points o Position within the current time segment and within the previous time segment, o _ t and o _ t 'Respectively o _ loc and o _ loc 'The corresponding timestamp, sp is the position variable, d ( sp,s _ loc )express sp and s _ The distance between d ( o _ _ )express o _ and s _ The distance between d ( _ ')express and o _ 'The distance between;
[0021] (3.2) Calculate the loss function of the seed point based on the evolutionary clustering smoothing operation L :
[0022] ;
[0023] in: is the distance threshold, d ( )express and The distance between β is a fixed parameter, a is a variable and a The value range includes and interval Positive integers within ( ) / ;
[0024] (3.3) With loss function L Minimize to determine the variables a , and then a Substitute into the loss function L Get the loss function value corresponding to the seed point;
[0025] (3.4) Let each moving object in the group be traversed as a seed point, and obtain the loss function value corresponding to each seed point according to steps (3.2) to (3.3). Take the seed point with the smallest loss function value as the optimal seed point of the group, and then determine its loss function. L Variables in a , and calculate the smoothed position of non-seed points in the group using the following expression;
[0026] o '( x )= a ( o ( x )- s ( x )) / + s ( x )
[0027] o '( y )= a ( o ( y )- s ( y )) / + s ( y )
[0028] in: o ( x )and o (y ) are non-seed points o The horizontal and vertical coordinates of the position before smoothing, o '( x )and o '( y ) are non-seed points o The horizontal and vertical coordinates of the smoothed position, s ( x )and s ( y ) are the horizontal and vertical coordinates of the optimal seed point position.
[0029] Furthermore, the specific implementation of step (4) is as follows: for any moving object in the current time segment p , according to the two-dimensional plane coordinates of the smoothed position, find the p Distance less than (generally greater than ) of all moving objects, if the number of these moving objects exceeds (generally greater than ), then p As the core points of clustering, the moving objects in the current time segment are clustered based on density according to these core points. is the distance threshold, is the quantity threshold.
[0030] Preferably, for any moving object in the current time segment p , according to its two-dimensional plane coordinates ( x , y )Establish p Grid index ,in x min and y min They are p The lower bounds of the horizontal and vertical coordinates of the plane, is the distance threshold; during the search process, only the grid index is searched moving objects, calculate the relationship between these moving objects and p The distance, where i = -1, 0 or 1, j = -1, 0, or 1; can speed up the construction of minimal groups and density-based clustering.
[0031] Preferably, in step (4), after completing a clustering, the modularity is calculated based on the clustering result. The greater the modularity, the higher the quality of the clustering. After fine-tuning, cluster again. If the modularity of the clustering result is improved, the adjusted distance threshold Used for clustering moving objects in the next time segment.
[0032] An evolutionary clustering system for moving objects, comprising:
[0033] An acquisition module is used to obtain the motion state information of the moving object online. Each piece of motion state information includes the geographical coordinates and timestamp of the moving object, and to determine the moving object within the current time segment.
[0034] Construct a module for converting the geographic coordinates of a moving object into two-dimensional plane coordinates based on a distance threshold As the length of the hypotenuse of the grid, the grid index of the moving object is constructed;
[0035] A grouping module is used to divide the moving objects in the current time segment into multiple minimized groups;
[0036] The smoothing module smoothes the positions of non-seed points in the group based on minimizing the seed points of the group;
[0037] Clustering module, used to cluster moving objects in the current time segment according to the smoothed positions;
[0038] Optimization module, used to optimize the distance threshold according to the modularity of the clustering results .
[0039] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor is used to execute the computer program to implement the above-mentioned evolutionary clustering method for moving objects.
[0040] A computer-readable storage medium stores a computer program, which is used to implement the above-mentioned evolutionary clustering method for moving objects when executed by a processor.
[0041] The present invention utilizes a minimized group structure to construct an efficient evolutionary clustering method for moving objects. Using grid indexing technology, it enables rapid searches for moving objects around a moving object, accelerating the generation of minimized groups and the execution of a density-based clustering algorithm. Using an evolutionary clustering smoothing algorithm and an evolutionary clustering loss function, it implements an algorithm that can calculate the optimal smoothed position of non-seed points in a group in approximately linear time, enabling efficient execution and applicability in real-time scenarios. Furthermore, the present invention employs an incremental density-based clustering algorithm to rapidly calculate the corresponding clustering results under different density clustering parameters, thereby optimizing clustering parameters in real-time scenarios. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] Schematic diagram of the flow of the evolutionary clustering method for moving objects of the present invention.
[0043] for Detailed expansion flow diagram of step S14 in FIG.
[0044] Schematic diagram of the smoothing algorithm based on evolutionary clustering.
[0045] Schematic diagram of density-based clustering.
[0046] for Detailed expansion flow diagram of step S15 in FIG.
[0047] This is a structural block diagram of the evolutionary clustering system for moving objects of the present invention.
[0048] The figure is a schematic diagram of the structure of an electronic device implemented according to the method of the present invention. DETAILED DESCRIPTION
[0049] In order to describe the present invention more specifically, the technical solution of the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.
[0050] like As shown, the evolutionary clustering method for moving objects of the present invention includes the following steps:
[0051] Step S11: obtaining the motion state information of the moving object online, where each piece of information includes the geographic location coordinates of the moving object and the current time point of the moving object.
[0052] There are a large number of moving objects in the display scene, and the sensors in the moving objects will record the geographical location of the moving objects at a certain moment and the time corresponding to this geographical location. A two-tuple (position, time) can be used to represent the motion state information of a moving object. Since we generally use longitude and latitude to represent the specific position of a moving object, in more common cases, this embodiment uses a three-tuple (longitude, latitude, time) to represent the motion state information of a moving object.
[0053] Step S12: Process the motion state information of the moving object online, perform Miller projection transformation and construct a grid index.
[0054] We transform the longitude and latitude in the motion state information of the moving object into a two-dimensional plane coordinate ( x ,y ), according to the plane position of the moving object, determine the lower bounds of the horizontal and vertical coordinates of the plane ( x min , y min ), and then based on the distance threshold of density clustering , construct the grid index of this moving object in the current plane as .
[0055] Step S13: For the moving objects within a time slice, a minimum group is constructed based on the set parameters and the grid index. Specifically:
[0056] For a moving object p , located at coordinates ( x , y ), assuming that the distance threshold of the group is minimized < , then we only need to search for the grid index moving objects, i = -1, 0 or 1, j =-1, 0 or 1; if the moving object p There are more than Spatial distance and p Less than moving object, then the moving object p A minimization group can be formed and become the seed point of this minimization group, and those spatial distances are p Less than The moving objects are divided into this minimized group.
[0057] Step S14: Based on the seed points in each minimization group, the non-seed points in the minimization group are subjected to a smoothing method based on evolutionary clustering. The specific process is as follows: As shown:
[0058] S41: Traverse the moving objects in a minimization group and assume that the moving object is smoothed as the seed point of the minimization group. Specifically: because the selection of seed points is random, for a minimization group, the current seed point is not necessarily the optimal seed point for smoothing calculation for this minimization group. Therefore, by traversing the moving objects in the minimization group and calculating the value of the smoothing loss function corresponding to each moving object as a seed point, the optimal seed point is selected. At the same time, in order to meet the speed limit constraint, the speed of the moving object currently selected as the seed point needs to be processed so that its speed meets d ( s _ _ ') / ( s _ t - s _ t ')< ,in s _ and s _ 'represents the position of the seed point in the current time segment and in the previous time segment, respectively. s _ t and s _ t 'Respectively s _ and s _ 'The corresponding timestamp, d ( s _ _ ')express s _ and s _ 'The distance between The maximum speed limit.
[0059] S42: For a given seed point, velocity preprocessing is first performed on other moving objects in the minimization group to satisfy the velocity constraint. Specifically, for moving objects that are not seed points in the minimization group, we need to perform velocity preprocessing on them before aligning and smoothing. When performing velocity preprocessing, it is necessary to ensure that the change in the distribution of the original data itself is sufficiently small and that the impact of velocity preprocessing on the smoothing calculation of the moving object is small. Therefore, when processing moving objects that are not seed points, the following relationship is used:
[0060] ;
[0061] d ( _ ') / ( o _ t - o _ t ')<
[0062] in: o _ and o _ 'represents the position of the non-seed point in the current time segment and in the previous time segment, respectively. o _ t and o _ t 'Respectivelyo _ and o _ 'The corresponding timestamp, is the position variable;
[0063] S43: Calculate the optimal smooth position of non-seed points according to the loss function of the smoothing method of evolutionary clustering. Specifically: for a non-seed point o And a given seed point s , the non-seed points can be calculated by the following loss function o The optimal smooth position adj(o):
[0064] ;
[0065] in: β is a fixed parameter, a is a variable and a The value range includes and interval Positive integers within ( ) / ;
[0066] Since the variable of the loss function a There are only a finite number of discrete values, and the loss function has a quadratic form, so for each non-seed point o and a given seed point s , adj(o) can be calculated with O(1) time complexity. As shown, o 2 After speed preprocessing, its position is adjusted to o 2. , then through the seed point o 1 is smoothed to adj( o 2).
[0067] S44: For a minimization group, select the moving object with the smallest loss function value when serving as a seed point, and set this moving object as the optimal seed point of the minimization group. Use this optimal seed point to smooth other moving objects in the minimization group until all minimization groups complete the smoothing calculation.
[0068] Specifically, for each minimization group, the moving object with the smallest loss function value obtained after traversal is selected as the optimal seed point of the minimization group, and the optimal seed point is used to smooth the other moving objects in the minimization group. The smoothing calculations between each minimization group are independent of each other and there is no interference. This is because for a moving object that is not a seed point, the loss of smoothing this moving object only depends on the position of the moving object itself. o _ and the position of the seed point of this minimization group s _ Because different minimization groups have completely different seed points, the loss calculations between different minimization groups are independent of each other. By minimizing the loss function value within each minimization group, a global minimum loss function value can be obtained, thereby obtaining the global optimal smooth position of all moving objects. The specific smoothing formula is as follows:
[0069] o '( x )= a ( o ( x )- s ( x )) / + s ( x )
[0070] o '( y )= a ( o ( y )- s ( y )) / + s ( y )
[0071] in: o ( x )and o ( y ) are non-seed points o The horizontal and vertical coordinates of the position before smoothing, o '( x )and o '( y ) are non-seed points o The horizontal and vertical coordinates of the smoothed position, s ( x )and s ( y ) are the horizontal and vertical coordinates of the optimal seed point position.
[0072] Step S15: Using grid index combined with density-based clustering method, the moving objects in a time slice are quickly clustered according to the set density-based clustering distance threshold and the minimum number of neighbors of the core point based on density clustering.
[0073] Based on density clustering, the surrounding area of the moving object is searched for the distance less than Is the number of moving objects greater than , to determine whether a moving object can become a core point in density-based clustering. The distance around a core point is less than The moving object will be in the same cluster as the core point; if the surrounding range of a core point is less than If there are other core points in the moving objects, then these two core points and all the moving objects in their clusters will be merged into one cluster. The density-based clustering continuously absorbs and merges the core points to form a cluster. The specific process is as follows: As shown:
[0074] S51: Use grid index combined with density-based clustering method for moving objects within a time slice, according to the set distance threshold of density-based clustering and the minimum number of neighbors of core points based on density clustering. Specifically: Using the constructed minimized group and grid index, we can reduce the search range of objects around a moving object, thereby greatly reducing the amount of calculation and quickly selecting core points from all moving objects.
[0075] S52: Using the core points calculated in S51, with the help of two data structures, hash table and queue, quickly cluster the moving objects within a time slice. Specifically: use the queue data structure to store the moving objects that are judged to be core points. Insert a core point into the queue each time it is processed, then take this core point out from the head of the queue, mark this core point as processed in the hash table, and cluster the objects with a distance less than 0. The core points that are not marked in the hash table are pushed into the queue until the queue is empty and a cluster is formed; then an unprocessed core point is taken out from the hash table and pushed into the queue, and the previous steps are repeated to continuously generate new clusters until all the seed points in the hash table are marked, and the clustering algorithm ends.
[0076] like As shown, there are 9 different moving objects o 1… o 9. Assumption =5, according to the diagram we can get o 1. o 3. o 7 is the core point, because the core pointo 1. o 7 and o The distance between 3 is less than ,therefore All moving objects in the cluster are in one cluster.
[0077] Step S16: Calculate the optimization direction of the distance threshold based on density clustering through the incremental density clustering algorithm to improve the quality of clustering. Specifically: calculate the modularity based on the clustering results. ,by +Δ and -Δ Perform incremental density-based clustering to calculate the corresponding _ and _ ,Compare 、 _ and _ , take out the maximum value, the maximum value corresponds to the next time slice should be optimized .
[0078] Corresponding to the aforementioned embodiment of the evolutionary clustering method for moving objects, the present invention also provides an embodiment of an evolutionary clustering system for moving objects.
[0079] like As shown, the evolutionary clustering system for moving objects in this embodiment includes:
[0080] An acquisition module 21 acquires the motion state information of a moving object online. Each piece of information includes the geographic coordinates of the moving object and the current time point of the moving object, and is divided into time slices.
[0081] The construction module 22 is used to convert the longitude and latitude coordinates into plane coordinates. According to the geographical location coordinates in the motion state information of the moving object within a time slice, the distance threshold based on density clustering is used. As the hypotenuse length of the grid, construct the grid index;
[0082] The group module 23 is used to generate a minimized group and seed points according to set parameters for a moving object within a time slice;
[0083] a smoothing module 24 for performing smoothing calculation on the generated minimization group to calculate the optimal smoothing position of each moving object in the minimization group;
[0084] A clustering module 25 is used to perform density-based clustering calculation on moving objects within a time slice by minimizing groups and grid index-assisted acceleration;
[0085] The optimization module 26 is configured to calculate optimized parameter values based on density clustering by using an incremental density clustering algorithm according to the clustering results of the moving objects within a time slice.
[0086] Regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.
[0087] Accordingly, the present invention also provides an electronic device, comprising a processor and a memory, wherein the memory is used to store a computer program, and the processor is used to execute the computer program to implement the above-mentioned evolutionary clustering method for moving objects. The figure shows a hardware structure of any electronic device with data processing capability where the evolutionary clustering system for moving objects provided by the embodiment of the present invention is located. In addition to the processor, memory, and network interface shown, the electronic device with data processing capabilities in which the system is located in the embodiment may also include other hardware according to the actual functions of the electronic device, which will not be described in detail.
[0088] Accordingly, the present invention also provides a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the above-described evolutionary clustering method for moving objects. The computer-readable storage medium can be an internal storage unit of the aforementioned electronic device with data processing capabilities, such as a hard disk or memory, or can be another external storage device, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc. equipped on the device. The computer-readable storage medium can also include both an internal storage unit of the electronic device with data processing capabilities and an external storage device, for storing the computer program and other programs and data required by the electronic device with data processing capabilities, and can also be used to temporarily store data that has been output or is about to be output.
[0089] The above description of the embodiments is intended to facilitate understanding and application of the present invention by those skilled in the art. It is apparent that those skilled in the art can readily make various modifications to the above embodiments and apply the general principles described herein to other embodiments without requiring creative effort. Therefore, the present invention is not limited to the above embodiments. Any improvements or modifications made by those skilled in the art based on the disclosure of the present invention should fall within the scope of protection of the present invention.
Claims
1. An evolutionary clustering method for moving objects, comprising the following steps: (1) obtaining the motion state information of the moving object online, each piece of motion state information including the geographic location coordinates and timestamp of the moving object; (2) Determine the moving objects in the current time segment and divide these moving objects into multiple minimization groups. Specifically: for any moving object p in the current time segment, transform the geographical location coordinates in its motion state information into two-dimensional plane coordinates (x, y) through the Miller projection method, and search to find all moving objects whose distance to p is less than md_thres. If the number of these moving objects exceeds mn_thres, form a minimization group with p as the initial seed point and divide all these moving objects into this group. md_thres is the distance threshold and mn_thres is the number threshold. (3) Based on the seed point of the minimized group, the positions of the non-seed points in the group are smoothed. The specific implementation method is as follows: 3.1 Take any moving object in the group as a seed point, and the other moving objects as non-seed points. Speed limit processing is performed on the seed points and non-seed points respectively. Specifically: For the seed point, s_loc is adjusted so that d(s_loc,s_loc') / (s_t-s_t')<maxSpeed, where s_loc and s_loc' represent the positions of the seed point in the current time segment and the previous time segment respectively, s_t and s_t' represent the timestamps corresponding to s_loc and s_loc' respectively, d(s_loc,s_loc') represents the distance between s_loc and s_loc', and maxSpeed is the maximum speed limit; For any non-seed point o, adjust o_loc to satisfy the following relationship: d(sp,o_loc') / (o_t-o_t')<maxSpeed in: o_loc and o_loc' respectively represent the positions of the non-seed point o in the current time segment and in the previous time segment, o_t and o_t' respectively represent the timestamps corresponding to o_loc and o_loc', sp is the position variable, d(sp,s_loc) represents the distance between sp and s_loc, d(o_loc,s_loc) represents the distance between o_loc and s_loc, d(sp,o_loc') represents the distance between sp and o_loc'; 3.2 Calculate the loss function L of the seed point based on the evolutionary clustering smoothing operation: Among them: md_thres is the distance threshold, represents rounding up, d(o_loc,s_loc) represents the distance between o_loc and s_loc, β is a fixed parameter, a is a variable and the range of a includes And interval A positive integer within Indicates rounding down; 3.3 Determine the variable a by minimizing the loss function L, and then substitute a into the loss function L to obtain the loss function value corresponding to the seed point; 3.4 Take each moving object in the group as a seed point, and obtain the loss function value corresponding to each seed point according to steps 3.2 to 3.
3. Take the seed point with the smallest loss function value as the optimal seed point of the group, and then determine the variable a in its loss function L, and calculate the smoothed position of the non-seed point in the group through the following expression; Where: o(x) and o(y) are the horizontal and vertical coordinates of the position of the non-seed point o before smoothing, o'(x) and o'(y) are the horizontal and vertical coordinates of the position of the non-seed point o after smoothing, s(x) and s(y) are the horizontal and vertical coordinates of the position of the optimal seed point; (4) Cluster the moving objects in the current time segment according to the smoothed positions. Specifically, for any moving object p in the current time segment, all moving objects whose distance to p is less than dd_thres are found by searching according to the two-dimensional plane coordinates of the smoothed position. If the number of these moving objects exceeds dn_thres, p is used as the core point of the cluster. Then, the moving objects in the current time segment are clustered based on density according to these core points. dd_thres is the distance threshold and dn_thres is the number threshold. For any moving object p in the current time segment, establish the grid index of p according to its two-dimensional plane coordinates (x, y) where x min and min are the lower bounds of the horizontal and vertical coordinates of the plane where p is located, and dd_thres is the distance threshold. During the search process, only the grid index is searched. , and calculate the distance between these moving objects and p, where i = -1, 0 or 1, j = -1, 0 or 1; After completing a clustering, the modularity is calculated based on the clustering results. The larger the modularity, the higher the quality of clustering. After fine-tuning the distance threshold dd_thres, clustering is performed again. If the modularity of the clustering result is improved, the adjusted distance threshold dd_thres is used for clustering moving objects in the next time segment.
2. The evolutionary clustering method for moving objects according to claim 1, characterized in that: In the step (2), when any motion status information is obtained online, if the difference between the timestamp of the motion status information and the timestamp of the first motion status information obtained in the current time segment is less than the set time interval, the moving object corresponding to the motion status information is classified as a moving object in the current time segment.
3. An evolutionary clustering system for moving objects, executing the evolutionary clustering method for moving objects as claimed in claim 1 or 2, characterized in that: The evolutionary clustering system comprises: An acquisition module is used to acquire the motion state information of the moving object online, each piece of motion state information includes the geographical location coordinates and timestamp of the moving object, and determine the moving object in the current time segment; A construction module is used to convert the geographic location coordinates of the moving object into two-dimensional plane coordinates, use the distance threshold dd_thres as the hypotenuse length of the grid, and construct the grid index of the moving object; A group module, used to divide the moving objects in the current time segment into multiple minimized groups; A smoothing module smoothes the positions of non-seed points in the group based on minimizing the seed points of the group; A clustering module is used to cluster moving objects in the current time segment according to the smoothed positions; The optimization module is used to optimize the distance threshold dd_thres according to the modularity of the clustering results.
4. A computer device comprising a memory and a processor, characterized in that: The memory stores a computer program, and the processor is used to execute the computer program to implement an evolutionary clustering method for moving objects as claimed in any one of claims 1 to 2.
5. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, it implements an evolutionary clustering method for moving objects as claimed in any one of claims 1 to 2.
