A trajectory anomaly detection method, system and product

Abstract

The present application relates to a kind of trajectory anomaly detection method, system and product.The method includes: obtaining start point coordinate and end point coordinate, and determining a group of trajectories according to the start point coordinate and the end point coordinate;A group of trajectories includes multiple trajectories;A trajectory is composed of a series of time-sequential coordinate points;Based on midpoint polar coordinate system, the latitude and longitude coordinates of the coordinate points in each trajectory are converted into polar coordinates;According to the polar coordinates, all the trajectories are converted into list-type features with equal length;Anomaly detector suitable for list-type features is used to detect anomalies in the list-type features, and determine abnormal trajectories.The present application improves the anomaly detection accuracy, and reduces time complexity and space complexity.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of abnormal trajectory detection, and in particular to a trajectory abnormality detection method, system and product. BACKGROUND

[0002] Trajectory data, also known as spatio-temporal data, is generated by positioning devices such as Global Positioning System (GPS), Wi-Fi, video surveillance or wireless sensor networks, and records the geographical information of moving objects. The moving objects can be people or animals, vehicles or aircraft, or even natural disasters such as hurricanes. Trajectory-based data mining techniques are mainly divided into four categories: prediction, clustering, classification and anomaly detection.

[0003] Currently, there is no uniform mathematical definition of "abnormal data". The most common definition is given by Hawkins: abnormal data is data that is very different from other observed data, leading people to suspect that it is generated by a different mechanism. Abnormal data may be noise, which is harmful to data analysis, and therefore should be removed before data analysis. The technology for identifying abnormal data is anomaly detection technology, which can be applied to fields such as financial fraud detection, fault detection, network intrusion detection and industrial system monitoring. As a subcategory of anomaly detection technology, abnormal trajectory detection technology has been used in various application scenarios, such as urban services, path exploration, destination prediction, public safety, sports behavior analysis, group behavior analysis, sports analysis, social services, ecological monitoring, environmental assessment and energy assessment.

[0004] Anomaly detection was first applied to trajectory data in 2000. Subsequently, anomaly techniques for other types of data were directly applied to trajectory data anomaly detection, such as Hidden Markov Models and clustering algorithms. Before applying these algorithms, mass data usually needs to be pre-processed, such as sampling or dimensionality reduction. However, these techniques introduced from other fields are not optimized for the abnormal trajectory detection task, so the accuracy is not high.

[0005] In November 2004, Qualcomm announced the success of mobile phone assisted GPS testing, and a large number of related technologies began to emerge, including abnormal trajectory detection technology. A typical abnormal trajectory detection process is as follows: first, pre-process the input data, then use related algorithms to score the data to quantify the abnormality of the data, then use post-processing techniques and integration techniques to optimize the abnormality score of the data, and finally determine whether the data is normal or abnormal according to the score threshold. The pre-processing part is particularly important for abnormal trajectory detection technology. Therefore, according to the pre-processing method used, abnormal trajectory detection technology can be divided into four categories: mapping-based, sampling-based, segmentation-based and other methods.

[0006] The pre-processing method based on mapping can be further divided into a grid mapping based method and a map mapping based method. The grid mapping based method grids the area where the trajectory is located, and represents the trajectory data as a series of grid points on the grid. Typical technologies based on this type of method include iBAT, iBOAT, DeepTEA, ATDC, etc. The advantage of this type of method is that it can avoid the influence of the sampling frequency of different trajectory recording devices and the speed of the object on the trajectory points, but the disadvantage is that it is sensitive to the grid size, which can cause two trajectories that are very close to each other to be represented by two completely different grid cells. The map mapping based method maps the trajectory data to the actual road traffic map, so additional map information is required. Typical technologies based on this type of method include RPat, DB-TOD, RNPAT, LoTAD, etc. The advantage of this type of method is that it reduces the influence of noise after mapping the trajectory to the real route, and the disadvantage is that the mapping process is very time-consuming, and cannot reflect the situation of new roads in the city, temporarily blocked roads and traffic jams.

[0007] The sampling based method obtains shorter or fixed length trajectories by sampling the original trajectory data, and can also obtain fixed length fragments by sliding window interception for trajectories in progress. Typical technologies based on this type of method include Piciarelli, RapidLOF, PN-Opt, STN-Outlier, etc. The disadvantage of this type of method is that due to the influence of the sampling frequency of the trajectory recording device and the speed of the object, the distribution of points on the trajectory is not necessarily uniform, which will cause two routes consistent trajectories to have a huge difference after sampling, thereby affecting the accuracy of the anomaly detection.

[0008] The segmentation based method divides the entire trajectory into several sub-trajectories, and identifies the sub-trajectory that contains an anomaly. The trajectory containing the abnormal sub-trajectory is also a global abnormal trajectory. Typical technologies based on this type of method include TRAOD, TODCSS, STC, TPBA, etc. Common segmentation methods include mutation point segmentation and artificial rule segmentation. The disadvantage of the former is that it is sensitive to noise and drifting trajectory points, and the disadvantage of the latter is that it requires manual experience to select parameters, thereby limiting the actual use.

[0009] As can be seen, the existing abnormal trajectory detection technology has some defects, mainly in the pre-processing stage, which directly affects the accuracy of the final anomaly detection, and has high time complexity and space complexity. SUMMARY

[0010] The purpose of the present application is to provide a trajectory anomaly detection method, system and product to solve the problem of low anomaly detection accuracy and high time complexity and space complexity.

[0011] To achieve the above purpose, the present application provides the following solutions:

[0012] A trajectory anomaly detection method, comprising:

[0013] obtaining a start point coordinate and an end point coordinate, and determining a set of trajectories according to the start point coordinate and the end point coordinate; a set of trajectories comprises a plurality of trajectories; a trajectory is composed of a series of time-sequenced coordinate points;

[0014] converting the longitude and latitude coordinates of the coordinate points in each of the trajectories into polar coordinates based on a midpoint polar coordinate system;

[0015] transforming all the trajectories into list-type features with equal lengths according to the polar coordinates;

[0016] adopting an anomaly detector suitable for list-type features to perform anomaly detection on the list-type features, and determining abnormal trajectories.

[0017] Optionally, the obtaining of the start point coordinate and the end point coordinate, and the determination of the set of trajectories according to the start point coordinate and the end point coordinate specifically comprise:

[0018] querying all trajectories similar to the start point coordinate and the end point coordinate from the historical trajectory database to determine a set of trajectories; wherein the similar trajectories are the start point coordinate of the similar trajectories, S is the start point coordinate, ||·|| is the Euclidean distance, and λ is the distance threshold, the end point coordinate of the similar trajectories, D is the end point coordinate.

[0019] Optionally, the conversion of the longitude and latitude coordinates of the coordinate points in each of the trajectories into polar coordinates based on the midpoint polar coordinate system specifically comprises:

[0020] determining a first line midpoint of the start point coordinate and the end point coordinate;

[0021] establishing a first polar coordinate system with the first line midpoint as a polar point and a direction from the first line midpoint to the start point coordinate as a polar axis;

[0022] converting the longitude and latitude coordinates of the coordinate points in each of the trajectories into polar coordinates based on the first polar coordinate system.

[0023] Optionally, the conversion of the longitude and latitude coordinates of the coordinate points in each of the trajectories into polar coordinates based on the midpoint polar coordinate system specifically comprises:

[0024] for each of the trajectories, determining a second line midpoint of the start point coordinate and the end point coordinate in any of the trajectories;

[0025] establishing a plurality of second polar coordinate systems with the midpoint of the second line as a pole and a direction from the midpoint of the second line to a starting point coordinate in the track as a polar axis, one of the second polar coordinate systems corresponding to one of the tracks;

[0026] converting latitude and longitude coordinates of any coordinate point in the track corresponding to the second polar coordinate system into polar coordinates based on any one of the second polar coordinate systems, until converting latitude and longitude coordinates of coordinate points in each of the tracks into polar coordinates.

[0027] Optionally, the converting latitude and longitude coordinates of coordinate points in each of the tracks into polar coordinates based on the midpoint polar coordinate system specifically includes:

[0028] determining a midpoint of a first line connecting the starting point coordinate and the ending point coordinate;

[0029] randomly selecting a track point from a group of tracks, establishing a third polar coordinate system with the midpoint of the first line as a pole and a direction from the midpoint of the first line to the track point as a polar axis;

[0030] converting latitude and longitude coordinates of coordinate points in each of the tracks into polar coordinates based on the polar coordinate system.

[0031] Optionally, the converting all the tracks into list-type features of equal length according to the polar coordinates specifically includes:

[0032] dividing a circumferential angle 2π into k parts, each part containing an arc range of 2π / k;

[0033] assigning an angular coordinate of each coordinate point on one of the tracks to any one of the k parts, and calculating an average value of radius coordinates of all coordinate points assigned to any one of the parts; the average value of the radius coordinates is a distance-based feature;

[0034] calculating a number of coordinate points assigned to any one of the parts; the number of coordinate points is a point number-based feature;

[0035] splicing the distance-based feature and the point number-based feature to determine a list-type feature, until converting all the tracks into list-type features of equal length.

[0036] Optionally, the converting all the tracks into list-type features of equal length according to the polar coordinates specifically includes:

[0037] dividing a circumferential angle π into k parts, each part containing an arc range of π / k;

[0038] assigning the angular coordinate of each coordinate point on the trajectory to any one of k parts, and calculating the average of the radius coordinates of all coordinate points assigned to any one part; the average of the radius coordinates is a distance-based feature;

[0039] counting the number of coordinate points with angular coordinates in [0, π) and the number of coordinate points with angular coordinates in [π, 2π); the number of coordinate points with angular coordinates in [0, π) and the number of coordinate points with angular coordinates in [π, 2π) are point number-based features;

[0040] concatenating the distance-based features and the point number-based features to determine list-type features until all the trajectories are converted into list-type features of equal length.

[0041] A trajectory anomaly detection system, comprising:

[0042] a set of trajectory determination modules configured to obtain start point coordinates and end point coordinates, and determine a set of trajectories based on the start point coordinates and the end point coordinates; a set of trajectories comprises a plurality of trajectories; a trajectory is composed of a series of coordinate points in time sequence;

[0043] a polar coordinate conversion module configured to convert the longitude and latitude coordinates of the coordinate points in each trajectory into polar coordinates based on a midpoint polar coordinate system;

[0044] a list-type feature conversion module configured to convert all the trajectories into list-type features of equal length based on the polar coordinates;

[0045] an anomaly detection module configured to perform anomaly detection on the list-type features using an anomaly detector suitable for list-type features, and determine abnormal trajectories.

[0046] An electronic device comprising a memory and a processor, the memory being configured to store a computer program, and the processor being configured to run the computer program to enable the electronic device to perform the trajectory anomaly detection method described above.

[0047] A computer-readable storage medium storing a computer program, the computer program being configured to be executed by a processor to implement the trajectory anomaly detection method described above.

[0048] According to the specific embodiments provided by the present application, the following technical effects are disclosed: the present application provides a trajectory anomaly detection method, system and product, coordinates of a group of trajectory points are converted into polar coordinates based on a midpoint polar coordinate system; all trajectories are converted into list-type features with equal lengths according to the polar coordinates; an anomaly detector suitable for list-type features is used to perform anomaly detection on the list-type features, determine abnormal trajectories, and is robust to various trajectory data conditions, does not rely on preprocessing techniques, and does not have the problem of low anomaly detection accuracy caused by the preprocessing stage, and all trajectories are converted into list-type features with equal lengths, with low time complexity and space complexity, and are more convenient for large-scale industrial applications. BRIEF DESCRIPTION OF DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0050] Figure 1 Flowchart of the trajectory anomaly detection method provided by the present application;

[0051] Figure 2 Process diagram of the trajectory data feature extraction based on the midpoint polar coordinate system provided by the present application;

[0052] Figure 3 Schematic diagram of three anomalies provided by the present application; Figure 3 (a) Schematic diagram of the anomaly type Detour provided by the present application; Figure 3 (b) Schematic diagram of the anomaly type Perturbation provided by the present application; Figure 3 (c) Schematic diagram of the anomaly type Route-switching provided by the present application;

[0053] Figure 4 Schematic diagram of three pairs of start-end point routes provided by the present application. DETAILED DESCRIPTION

[0054] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0055] The application aims to provide a trajectory anomaly detection method, system and product, improve anomaly detection accuracy, and reduce time complexity and space complexity.

[0056] In order to make the above-mentioned purposes, features and advantages of the application more obvious and easy to understand, the application will be further described in detail below in combination with the drawings and specific embodiments.

[0057] Embodiment one

[0058] Figure 1 The flowchart of the trajectory anomaly detection method provided by the application is shown in Figure 1 The trajectory anomaly detection method comprises the following steps.

[0059] Step 101: Obtain a starting point coordinate and an ending point coordinate, and determine a group of trajectories according to the starting point coordinate and the ending point coordinate; a group of trajectories comprises a plurality of trajectories; one trajectory is composed of a series of coordinate points in time sequence. The trajectory can be a GPS trajectory or a Beidou satellite navigation system trajectory, etc.

[0060] In actual application, the step 101 specifically comprises: querying all trajectories similar to the starting point coordinate and the ending point coordinate from the historical trajectory database to determine a group of trajectories; wherein the similar trajectories are The starting point coordinate of the similar trajectory is S, the starting point coordinate is S, ||·|| is the Euclidean distance, and λ is the distance threshold value. The ending point coordinate of the similar trajectory is D, and the ending point coordinate is D.

[0061] One trajectory is composed of a series of coordinate points in time sequence, that is Wherein, T j is the jth trajectory, i is a time point, is the coordinate of the ith time point of the jth trajectory, and are the dimension coordinate and the longitude coordinate respectively, is the coordinate of the mth time point of the jth trajectory.

[0062] Given a starting point coordinate S={x s ,y s} and an ending point coordinate D={x d ,y d}, first find all trajectories similar to the starting point and the ending point from the historical trajectory database, wherein the similarity is defined as In the above formula, ||·|| represents the Euclidean distance, and λ is the distance threshold value. Finally, a group of required trajectories G S,D,λ ={T1,T2,…,Tj ,…T n};T n Let n be the nth trajectory.

[0063] The objective of this invention is to identify abnormal trajectories in this set of trajectories, where abnormality is defined as a trajectory with a relatively small number of routes or a significant difference from other routes.

[0064] Step 102: Based on the midpoint polar coordinate system, convert the latitude and longitude coordinates of the coordinate points in each trajectory into polar coordinates.

[0065] In practical applications, the latitude and longitude coordinates of the coordinate points in the trajectory are represented as polar coordinates. This invention proposes three methods.

[0066] A1) Step 102 specifically includes: determining the midpoint of the first line connecting the starting point coordinates and the ending point coordinates; establishing a first polar coordinate system with the midpoint of the first line as the pole and the direction from the midpoint of the first line to the starting point coordinates as the polar axis; and converting the latitude and longitude coordinates of each coordinate point in the trajectory into polar coordinates based on the first polar coordinate system.

[0067] Find the midpoint M of the line connecting the starting point S and the ending point D, and take it as the pole. Establish a polar coordinate system for the polar axis. G S,D,λ Any coordinate point on any trajectory It can be represented in polar coordinates as Wherein, angular coordinates radius coordinates

[0068] A2) Step 102 specifically includes: for each trajectory, determining the midpoint of the second line connecting the starting point coordinates and the ending point coordinates in any trajectory; establishing multiple second polar coordinate systems with the midpoint of the second line as the pole and the direction from the midpoint of the second line to the starting point coordinates in the trajectory as the polar axis; one second polar coordinate system corresponds to one trajectory; based on any second polar coordinate system, converting the latitude and longitude coordinates of any coordinate point in the trajectory corresponding to the second polar coordinate system into polar coordinates, until the latitude and longitude coordinates of the coordinate points in each trajectory are converted into polar coordinates.

[0069] For G S,D,λ A trajectory in Find its own starting point and the end point Midpoint M of the line connecting j Taking it as the extreme point, Establish a polar coordinate system with the polar axis. Any coordinate point on this trajectory... It can be represented in polar coordinates as Wherein, angular coordinates radius coordinate

[0070] A3) the step 102 specifically comprises: determining the midpoint of the first connecting line of the start point coordinate and the end point coordinate; randomly selecting a trajectory point from a group of trajectories, establishing a third polar coordinate system with the midpoint of the first connecting line as the pole and the direction from the midpoint of the first connecting line to the trajectory point as the polar axis; and converting the longitude and latitude coordinates of the coordinate points in each of the trajectories into polar coordinates based on the polar coordinate system.

[0071] Find the midpoint M of the connecting line of the start point S and the end point D, and take it as the pole. Then randomly select a trajectory point Q from G S,D,λ , and establish a polar coordinate system with as the polar axis. S,D,λ Any coordinate point on any trajectory in G can be expressed in polar coordinates as , wherein

[0072] Step 103: converting all the trajectories into list-type features with equal lengths according to the polar coordinates.

[0073] In actual applications, the unequal-length trajectory data is extracted into equal-length list-type features, and the present application proposes two methods:

[0074] B1) the step 103 specifically comprises: dividing the circular angle 2π into k parts, each containing an arc range of 2π / k; assigning the angular coordinate of each coordinate point on a trajectory to any one of the k parts, and calculating the average value of the radius coordinates of all the coordinate points assigned to any one part; the average value of the radius coordinates is a distance-based feature; calculating the number of coordinate points assigned to any one part; the number of coordinate points is a point number-based feature; and splicing the distance-based feature and the point number-based feature to determine a list-type feature, until all the trajectories are converted into list-type features with equal lengths.

[0075] After all the coordinate points of a trajectory are expressed in the form of polar coordinates, the present application divides the circular angle 2π into k parts, i.e. each part contains an arc range of 2π / k, expressed as Then, according to the angular coordinate of each coordinate point on a trajectory, it is assigned to a certain part of the k parts, i.e.

[0076] After all the coordinate points are assigned, for all the coordinate points assigned to , the average value of their radius coordinates d is calculated and expressed as , i.e. Thus, we have The invention refers to this as the distance-based feature.

[0077] Then, the number of coordinate points assigned into is calculated and denoted as |·| represents the number of elements in the set, thus we have The invention refers to this as the point number-based feature.

[0078] Finally, the two features are concatenated to obtain the final feature, i.e.

[0079] B2) the step 103 specifically comprises: dividing the half circle angle π into k parts uniformly, each part containing an arc range of π / k; assigning the angular coordinate of each coordinate point on the trajectory (the angular coordinate exceeding π is changed to 2π minus the angular coordinate) into any one of the k parts, and calculating the average value of the radius coordinates of all coordinate points assigned into any one part; the average value of the radius coordinates is the distance-based feature; counting the number of coordinate points with the angular coordinate in [0, π) and the number of coordinate points with the angular coordinate in [π, 2π); the number of coordinate points with the angular coordinate in [0, π) and the number of coordinate points with the angular coordinate in [π, 2π) are the point number-based features; and concatenating the distance-based feature and the point number-based feature to determine the list-type feature, until all the trajectories are converted into list-type features with equal lengths.

[0080] Divide the half circle angle π into k parts uniformly, i.e. each part contains an arc range of π / k, denoted as Then, according to the angular coordinate of each coordinate point on the trajectory, it is assigned into a certain part of the k parts, i.e.

[0081] After all the coordinate points are assigned, for all the coordinate points assigned into , the average value of their radius coordinates d is calculated and denoted as i.e. Thus, we have The invention refers to this as the distance-based feature.

[0082] Then, the number of coordinate points with the angular coordinate in [0, π) is counted and recorded as The number of coordinate points with the angular coordinate in [π, 2π) is counted and recorded as Thus, we have The invention refers to this as the point number-based feature.

[0083] Finally, the two features are concatenated to obtain the final feature, i.e.

[0084] Ultimately, all trajectories G S,D,λ ={T1,T2,…,T j ,…T n} is converted into a feature list F = [F1, F2, ..., F j ,…F n ].

[0085] Step 104: Use an anomaly detector suitable for list-type features to perform anomaly detection on the list-type features and determine the trajectory of the anomaly.

[0086] In practical applications, any anomaly detector suitable for list-type features, such as the Local Anomaly Factor (LOF) method or the K-Nearest Neighbors (KNN) method, can be applied to the transformed features to detect abnormal trajectories.

[0087] In this invention, LOF is used as the anomaly detector.

[0088] Regarding steps 102-103, the present invention refers to the combination of method A1 and method B1 as GMiPo, the combination of method A1 and method B2 as MiPo, the combination of method A1 and only using the distance-based features of method B2 as MiPoD, the combination of method A1 and only using the point-based features of method B2 as MiPoP, and the combination of method A2 and method B2 as PMiPo.

[0089] First, this invention provides a simple example of MiPo. For example... Figure 2 As shown, MiPo extracts two sets of features from trajectory data: distance-based features. ( Figure 2 Top box B1) and point-based features ( Figure 2 (Bottom box B2), thus storing the trajectory data {T1,T2,T3,T4}. Figure 2 The data in the lower left box A) can be converted into table data [F1, F2, F3, F4]( Figure 2 (Bottom right box C). To extract distance-based features, MiPo uses midpoint polar coordinates to represent each point in the trajectory data, a vector. As a reference vector, M is the midpoint of the line connecting the starting point S and the destination point D. Then, the points in each trajectory can be divided into three parts, such as... Figure 2 As shown to the left of box B1. For example, the distribution of points in trajectory T1 is as follows: Then, the average distance between each point in each subset and M is calculated, yielding values ​​of 3.4, 3.0, and 2.4. Therefore, the distance-based feature of T1 is... Since all six points in T1 are located on one side of the line SD, i.e. their angles are all between 0 and π, the point number based feature of T1 is Finally, the present application splices the two features together as the final feature of T1, i.e. F1 = [3.4, 3.0, 2.4, 6, 0].

[0090] Secondly, the present application designs a series of perfect experiments based on a real GPS driving trajectory dataset to compare the present application with other abnormal trajectory detection techniques. The dataset contains 1710670 driving trajectories of 442 taxis in Porto, Portugal between January 7, 2013 and January 30, 2014, and the GPS sampling frequency is 1 / 15 Hz. The present application selects 3 pairs of start-end points from the dataset, and constructs 12 test sets by combining the forms of manually selecting trajectories, manually labeling abnormalities and manually adding abnormalities. Among them, there are three types of manually added abnormalities, which are detour, perturbation and route-switching. Detour represents that the driver detours a long distance in the normal route, perturbation represents that the driver often detours a small range in the process of driving on the normal route, and route-switching represents that the driver changes from one route to another normal route. Figure 3 The schematic diagram of the three abnormalities provided by the present application is shown in Figure 3 Table 1 is the information table of the twelve test sets, and the specific test set information is shown in Table 1.

[0091] Table 1

[0092]

[0093] In Table 1, the AiSq-prefixed dataset represents the route based on the first pair of start-end points, the StSt-prefixed dataset represents the route based on the second pair of start-end points, and the UnCh-prefixed dataset represents the route based on the third pair of start-end points. Figure 4 The schematic diagram of the three pairs of start-end points provided by the present application is shown in Figure 4 .

[0094] On the 12 test sets, the five embodiments of GMiPo, MiPo, MiPoD, MiPoP and PMiPo proposed in the application, and eight typical abnormal trajectory detection techniques of IBAT, IBOAT, RapidLOF, LSTM, ATDC, TPBA, ELSTMAE and RNPAT were tested and compared, and the area under receiver operating characteristic curve (AUC-ROC) commonly used in the field of anomaly detection was used as an evaluation index. The value of the index is between [0, 1], and the larger the value, the better the performance of the technology. Table 2 is the results of the five embodiments and eight comparative techniques on the twelve test sets, and the experimental results are shown in Table 2.

[0095] Table 2

[0096]

[0097] In Table 2, AVG, STD and Media respectively represent the average value, standard deviation and median of the results of a method on all test sets. As can be seen from Table 2, among all the proposed schemes, MiPo has the best performance, with an average AUC of 0.99. Compared with it, the average AUC of GMiPo is 0.96, and the performance is slightly lower, which shows that method B2 is better than method B1. The performance of MiPoP and MiPoD is lower than that of MiPo, which shows that the combination of distance-based feature domain and point-based feature is more advantageous. The performance of PMiPo is very close to that of MiPo, which shows that there is no obvious advantage or disadvantage between method A1 and method A2.

[0098] Among the other abnormal trajectory detection techniques compared, RNPAT needs to use the city road map as additional information to match the trajectory with the real road, so it is not applicable to the test set with artificially added abnormal routes, so it has no results for the first 8 data sets. The comparison results show that on all test sets, no matter which pair of start-end routes, which type of abnormal trajectory and abnormal ratio, MiPo is better than all other abnormal trajectory detection techniques. At the same time, the standard deviation of the results of MiPo is the smallest, which is 0.01, which means that it has robustness and stability.

[0099] Table 3

[0100]

[0101] Finally, the present application tests the operation time consumption of all technologies. Table 3 is the operation time test table of MiPo scheme correlation and 8 kinds of comparison technologies, and the results are shown in Table 3, the time unit is second, n represents the number of trajectories tested, and m represents the length of each trajectory. Mipo represents the time consumption of step 102+step 103, Mipo- represents the time consumption of step 103, MiPo+LOF represents the time consumption of step 102+step 103+step 104. The results show that in the case that the midpoint polar coordinates of the trajectories have been converted, the feature extraction of 10000 trajectories with a length of 90 only takes about 0.78 seconds, that is, only 1 microsecond is needed for each trajectory on average. If the time of midpoint polar coordinate conversion is included, it only takes 2.7 milliseconds on average to extract the features of each trajectory. This shows that the running speed of MiPo is very fast, and it can be further accelerated by converting the midpoint polar coordinates of the trajectories in advance and storing them in the database.

[0102] In the horizontal comparison, the speed of MiPo+LOF is almost 10 times that of IBAT, which is a technology with competitive performance. If only MiPo- and LOF are considered, it is even 100 times faster than IBAT. The fastest among the remaining technologies is RapidLOF, but as can be seen from the RLOF column of Table 2, its performance is quite different. Compared with the deep learning-based methods LSTM and ELSTMAE, the speed of MiPo+LOF is more than 40 times faster. And due to the preprocessing of mapping trajectories to road graphs, RNPAT is the most time-consuming.

[0103] Embodiment Two

[0104] In order to perform the method corresponding to the above-mentioned embodiment one, to realize the corresponding functions and technical effects, the following provides a trajectory anomaly detection system.

[0105] A trajectory anomaly detection system comprises:

[0106] A set of trajectory determination modules are configured to obtain a starting point coordinate and an ending point coordinate, and determine a set of trajectories based on the starting point coordinate and the ending point coordinate; a set of trajectories comprises a plurality of trajectories; and a trajectory is composed of a series of coordinate points in time sequence.

[0107] A polar coordinate conversion module is configured to convert the latitude and longitude coordinates of the coordinate points in each trajectory into polar coordinates based on a midpoint polar coordinate system.

[0108] A list-type feature conversion module is configured to convert all the trajectories into list-type features with equal lengths based on the polar coordinates.

[0109] An anomaly detection module is configured to adopt an anomaly detector suitable for list-type features to perform anomaly detection on the list-type features, and determine abnormal trajectories.

[0110] Embodiment three

[0111] The electronic device provided by the embodiment of the present application comprises a memory and a processor. The memory is used to store a computer program. The processor executes the computer program to make the electronic device execute the trajectory anomaly detection method of the embodiment one.

[0112] In practical application, the electronic device can be a server.

[0113] In practical application, the electronic device comprises at least one processor, a memory, a bus and a communications interface.

[0114] The processor, the communications interface and the memory complete mutual communication through the communications bus.

[0115] The communications interface is used to communicate with other devices.

[0116] The processor is used to execute a program, and specifically can execute the method described in the above embodiments.

[0117] Specifically, the program can comprise program codes, and the program codes comprise computer operation instructions.

[0118] The processor can be a central processing unit (CPU) or an application specific integrated circuit (ASIC) or one or more integrated circuits configured to implement the embodiments of the present application. The one or more processors included in the electronic device can be processors of the same type, such as one or more CPUs; or can be processors of different types, such as one or more CPUs and one or more ASICs.

[0119] The memory is used to store the program. The memory can comprise a high-speed RAM memory, and can also comprise a non-volatile memory, such as at least one disk memory.

[0120] Based on the description of the above embodiments, the embodiment of the present application provides a storage medium having computer program instructions stored thereon. The computer program instructions can be executed by a processor to implement the method described in any embodiment

[0121] The trajectory anomaly detection system provided by the embodiment of the present application exists in various forms, including but not limited to:

[0122] (1) Mobile communication device: This type of device is characterized by having mobile communication function and providing voice and data communication as the main target. This type of terminal includes smart phone (e.g. iPhone), multimedia phone, functional phone, and low-end phone, etc.

[0123] (2) Ultra-mobile personal computer device: This type of device belongs to the category of personal computer, having computing and processing function, and generally having mobile Internet performance. This type of terminal includes PDA, MID and UMPC device, such as iPad.

[0124] (3) Portable entertainment device: This type of device can display and play multimedia content. This type of device includes audio and video player (e.g. iPod), hand-held game console, electronic book, and smart toy and portable car navigation device.

[0125] (4) Other electronic devices with data interaction function.

[0126] So far, specific embodiments of the present subject matter have been described. Other embodiments are within the scope of the following claims. In some cases, the actions noted in the claims can occur in a different order and still achieve the desired results. Also, the processes depicted in the figures do not necessarily require the particular order shown, or sequential order, to achieve the desired results. In certain implementations, multitasking and parallel processing can be advantageous.

[0127] The systems, apparatuses, modules or units illustrated by the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer can be a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0128] For the convenience of description, the above apparatuses are described by dividing them into various units with functions respectively. Of course, the functions of the units can be implemented in one or more software and / or hardware in implementing the present application. Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0129] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or combination thereof. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or combination thereof.

[0130] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or combination thereof. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or combination thereof.

[0131] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or combination thereof. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or combination thereof.

[0132] In one typical configuration, the computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0133] The memory can include non-persistent memory and / or volatile memory, such as random access memory (RAM) and / or cache memory, non-volatile memory, such as read-only memory (ROM), EPROM, and / or flash memory. The memory is an example of computer-readable media.

[0134] Computer-readable media, including both permanent and non-permanent, removable and non-removable media, can store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, and CD-ROM.

[0135] Digital multifunction optical disc (DVD) or other optical storage, magnetic cassette tape, magnetic magnetic disk storage or other magnetic storage devices

[0136] Or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transient media, such as modulated data signals and carrier waves.

[0137] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0138] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0139] This application can be described in the general context of computer-executable instructions that are executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific transactions or implement specific abstract data types. This application can also be practiced in distributed computing environments where transactions are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0140] The various embodiments described in this specification are presented for the purpose of illustrating the principles of the present application and its best mode of operation. Each of the embodiments described in this specification has been provided for the purpose of illustration and not limitation. The same or similar parts of the various embodiments are referenced by the same or similar reference numerals.

[0141] The principles and operation of the present application are explained so as to enable any person skilled in the art to make and use the present application. While the present application has been described in relation to particular embodiments thereof, many other possible variations, modifications, and applications thereof will come to mind to those skilled in the art upon the reading and understanding of the foregoing description. Such modifications to the inventive concept as come within the scope of the appended claims are meant to be within the scope of the present application.

Claims

1. A trajectory anomaly detection method, characterized by, The method comprises the following steps: acquiring a start point coordinate and an end point coordinate, and determining a set of trajectories based on the start point coordinate and the end point coordinate; a set of trajectories comprises a plurality of trajectories; a trajectory is composed of a series of coordinate points in time sequence; converting the latitude and longitude coordinates of the coordinate points in each trajectory into polar coordinates based on a midpoint polar coordinate system; transforming all the trajectories into list-type features with equal lengths according to the polar coordinates; performing anomaly detection on the list-type features by using an anomaly detector suitable for list-type features, to determine abnormal trajectories; the step of transforming all the trajectories into list-type features with equal lengths according to the polar coordinates specifically comprises: dividing a circumferential angle 2π into k parts, each part containing an arc range of 2π / k; assigning the angular coordinate of each coordinate point on a trajectory to any one of the k parts, and calculating the average value of the radius coordinates of all the coordinate points assigned to any one part; the average value of the radius coordinates is a distance-based feature; calculating the number of coordinate points assigned to any one part; the number of coordinate points is a point number-based feature; splicing the distance-based feature and the point number-based feature to determine a list-type feature, until all the trajectories are transformed into list-type features with equal lengths; or dividing a circumferential angle π into k parts, each part containing an arc range of π / k; assigning the angular coordinate of each coordinate point on a trajectory to any one of the k parts, and calculating the average value of the radius coordinates of all the coordinate points assigned to any one part; the average value of the radius coordinates is a distance-based feature; counting the number of coordinate points with angular coordinates in [0, π) and the number of coordinate points with angular coordinates in [π, 2π); the number of coordinate points with angular coordinates in [0, π) and the number of coordinate points with angular coordinates in [π, 2π) are point number-based features; splicing the distance-based feature and the point number-based feature to determine a list-type feature, until all the trajectories are transformed into list-type features with equal lengths. 2.The trajectory anomaly detection method of claim 1, wherein, the step of acquiring a start point coordinate and an end point coordinate, and determining a set of trajectories based on the start point coordinate and the end point coordinate specifically comprises: From the historical trajectory database, query all trajectories similar to the start point coordinates and the end point coordinates to determine a set of trajectories; wherein the similar trajectories are , the start point coordinates of the similar trajectories, and S is the start point coordinates, the Euclidean distance, and λ is the distance threshold, the end point coordinates of the similar trajectories, and D is the end point coordinates. 3.The trajectory anomaly detection method of claim 2, wherein, the step of converting the latitude and longitude coordinates of the coordinate points in each trajectory into polar coordinates based on a midpoint polar coordinate system specifically comprises: determining the midpoint of a first line connecting the start point coordinate and the end point coordinate; establishing a first polar coordinate system with the midpoint of the first line as the polar point and the direction from the midpoint of the first line to the start point coordinate as the polar axis; converting the latitude and longitude coordinates of the coordinate points in each trajectory into polar coordinates based on the first polar coordinate system. 4.The trajectory anomaly detection method of claim 2, wherein, the step of converting the latitude and longitude coordinates of the coordinate points in each trajectory into polar coordinates based on a midpoint polar coordinate system specifically comprises: for each trajectory, determining the midpoint of a second line connecting the start point coordinate and the end point coordinate in any trajectory; establishing a plurality of second polar coordinate systems with the midpoint of the second line as the polar point and the direction from the midpoint of the second line to the start point coordinate in the trajectory as the polar axis; one second polar coordinate system corresponds to one trajectory; Converting the latitude and longitude coordinates of any coordinate point in the trajectory corresponding to the second polar coordinate system into polar coordinates based on any of the second polar coordinate systems, until the latitude and longitude coordinates of the coordinate points in each of the trajectories are converted into polar coordinates.

5. The trajectory anomaly detection method of claim 2, wherein, The latitude and longitude coordinates of the coordinate points in each of the trajectories are converted into polar coordinates based on the midpoint polar coordinate system, specifically including: Determining the midpoint of the first line connecting the start point coordinate and the end point coordinate; Randomly selecting a trajectory point from a group of trajectories, establishing a third polar coordinate system with the midpoint of the first line as the polar point and the direction from the midpoint of the first line to the trajectory point as the polar axis; Converting the latitude and longitude coordinates of the coordinate points in each of the trajectories into polar coordinates based on the polar coordinate system.

6. A trajectory anomaly detection system characterized by, Including: A group of trajectory determination module for obtaining start point coordinates and end point coordinates, and determining a group of trajectories according to the start point coordinates and the end point coordinates; a group of trajectories includes multiple trajectories; a trajectory is composed of a series of time-ordered coordinate points; The latitude and longitude coordinates of the coordinate points in each of the trajectories are converted into polar coordinates based on the midpoint polar coordinate system; The list type feature conversion module is used for converting all the trajectories into list type features with equal length according to the polar coordinates, specifically including: dividing the circular angle 2π into k parts, each part containing an arc range of 2π / k; assigning the angular coordinate of each coordinate point on a trajectory to any one of the k parts, and calculating the average value of the radius coordinates of all coordinate points assigned to any one part; the average value of the radius coordinates is a distance-based feature; calculating the number of coordinate points assigned to any one part; the number of coordinate points is a point-based feature; splicing the distance-based feature and the point-based feature to determine the list type feature, until all the trajectories are converted into list type features with equal length; or dividing the half circular angle π into k parts, each part containing an arc range of π / k; assigning the angular coordinate of each coordinate point on a trajectory to any one of the k parts, and calculating the average value of the radius coordinates of all coordinate points assigned to any one part; the average value of the radius coordinates is a distance-based feature; counting the number of coordinate points with angular coordinates in [0, π) and the number of coordinate points with angular coordinates in [π, 2π); the number of coordinate points with angular coordinates in [0, π) and the number of coordinate points with angular coordinates in [π, 2π) are point-based features; splicing the distance-based feature and the point-based feature to determine the list type feature, until all the trajectories are converted into list type features with equal length; The abnormality detection module is used for adopting an abnormality detector suitable for list type features to perform abnormality detection on the list type features, and determining abnormal trajectories.

7. An electronic device, comprising: The electronic device includes a memory for storing a computer program and a processor for running the computer program to enable the electronic device to perform the trajectory abnormality detection method of any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The computer program is stored in the computer storage and is executed by the processor to implement the trajectory anomaly detection method according to any one of claims 1-5.

Images