An airport scene semantic trajectory representation and taxiing mode anomaly detection method

Abstract

The application relates to an airport scene semantic trajectory representation and taxiing mode anomaly detection method, belongs to the field of airport operation management, realizes semantic conversion based on aircraft motion characteristics in combination with scene map topological information, and obtains an airport scene aircraft semantic trajectory representation; on the basis, a flight path similarity matrix is constructed by using a road section geographic information to distribute weights and improving a longest common subsequence method, and scene taxiing mode recognition is realized by using hierarchical clustering; finally, a maximum outlier value for judging mode anomaly is obtained by calculating the internal similarity of the recognized mode, similarity of subsequent input flight paths and each mode is calculated, anomaly detection is realized by comparison of the maximum outlier value, abnormal flight paths are early warned, scene scheduling safety is maintained, and scene scheduling efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of airport operation management, and specifically to a method for airport surface semantic trajectory representation and taxiing pattern anomaly detection. Background Technology

[0002] The sustained and rapid development of the air transport industry has led to a rapid increase in airport surface traffic. The contradiction between the dramatic increase in surface traffic demand and its limited capacity has become increasingly prominent, making the improvement of airport operational efficiency and the detection of anomalies during taxiing critical issues that urgently need to be addressed in surface operations management. Many operations aimed at improving surface operational efficiency and detecting taxiing anomalies (e.g., determining aircraft deviations from taxiways, issuing warnings of runway incursions, and scheduling runway operations) rely on simplified representations of typical surface taxiing trajectories. While these tasks can be accomplished through clustering of taxiing trajectory points or manual specification, the significant number of missing and offset noise points in the original track sampling points, as well as potential accuracy errors in manual annotation, pose significant challenges that cannot be ignored.

[0003] In the prior art, semantic trajectory representation of surface aircraft and taxiing pattern anomaly detection have been proposed to solve the above problems. However, the semantic trajectory representation of aircraft and taxiing pattern anomaly detection have the following problems: (1) There are many missing original field monitoring track data and large offsets. The density of the airport surface road network is large and the structure is complex. The association process needs to overcome the noise problem and it is difficult to accurately obtain a semantic trajectory representation that conforms to the actual operation characteristics of the aircraft; (2) The semantic trajectory of the aircraft is represented in the form of map road segment number. How to calculate the similarity between tracks based on the geographical information implied by the number, that is, the probability of belonging to the same pattern, and identify the taxiing pattern contained in the surface track according to the accuracy requirements of pattern differentiation; (3) After obtaining the taxiing pattern differentiation between tracks, how to define abnormal taxiing patterns and realize pattern recognition and anomaly detection of the input track in real time. Summary of the Invention

[0004] In view of the above problems, this invention provides a method for airport surface semantic trajectory representation and taxiing pattern anomaly detection. Based on the dynamic characteristics of aircraft taxiing on the airport surface and combined with surface map topology information, semantic transformation is achieved to obtain the semantic trajectory representation of the aircraft on the airport surface. On this basis, a longer longest common subsequence method is constructed by allocating segment weights using segment geographic information, and hierarchical clustering is used to recognize surface taxiing patterns. Finally, the maximum outlier value for judging pattern anomalies is obtained by calculating the internal similarity of the identified patterns. The similarity of subsequent input tracks with each pattern is calculated, and anomaly detection is achieved by comparing the maximum outlier value, providing early warning of abnormal tracks, maintaining airport scheduling safety, and improving airport scheduling efficiency.

[0005] This invention provides a method for airport surface semantic trajectory representation and taxiing pattern anomaly detection, including:

[0006] Step 1: Obtain multiple initial taxiing trajectories of aircraft based on the surface surveillance radar system; obtain multiple aircraft taxiing trajectory data by matching the multiple initial taxiing trajectories of aircraft through map matching;

[0007] By utilizing multiple aircraft taxi trajectory data, multiple initial aircraft semantic trajectory representations are obtained;

[0008] Based on the division of parking positions, multiple initial aircraft semantic trajectory representations are cleaned and their integrity is verified to obtain multiple aircraft semantic trajectory representations;

[0009] Preferably, the specific steps for obtaining multiple initial aircraft semantic trajectory representations include:

[0010] Step 11: Under the constraints of aircraft semantic trajectory representation, obtain multiple initial taxiing trajectories of the aircraft based on the surface surveillance radar system;

[0011] It is understandable that the multiple aircraft taxiing trajectories are derived from takeoff / landing behavior;

[0012] Airport contour coordinates are established by map matching, and noisy trajectory points in multiple aircraft initial taxi trajectories are deleted using the airport contour coordinates to obtain multiple aircraft taxi trajectory data.

[0013] Based on the multiple aircraft taxiing trajectory data, a sequence of multiple aircraft taxiing trajectory points is generated;

[0014] Step 12: Establish a candidate road segment set by taking all road segments of the road network as candidate road segments; the first trajectory point in each aircraft taxiing trajectory point sequence starts from the runway end; each aircraft taxiing trajectory point sequence includes multiple runway ends and parking position ends;

[0015] Step 13: Select the c-th aircraft taxiing trajectory point sequence, where c = 1, 2, 3…C, and C represents the total number of aircraft taxiing trajectory point sequences;

[0016] Obtain the initial trajectory point c1 of the c-th aircraft taxiing trajectory point sequence, and obtain the minimum projection distance from the initial trajectory point c1 to each candidate road segment in the candidate road segment set using the Euclidean distance method, thus obtaining multiple projection distances corresponding to the initial trajectory point c1.

[0017] Step 14: Set the maximum error distance d for the aircraft semantic trajectory points. max The maximum error distance of the aircraft semantic trajectory point is compared and filtered with multiple projection distances corresponding to the initial trajectory point c1 to select multiple candidate projection distances for the initial trajectory point c1.

[0018] Sort the multiple candidate projection distances of the initial trajectory point c1 to obtain the sorted candidate projection distances; the smallest candidate projection distance among the sorted candidate projection distances is represented as the matching road segment of the initial trajectory point c1.

[0019] Furthermore, the multiple candidate projection distances of the initial trajectory point c1 are the maximum error distance d among the multiple projection distances corresponding to the initial trajectory point c1 that is less than the aircraft semantic trajectory point. max When, the candidate projection distance of the corresponding initial trajectory point.

[0020] Step 15: Let a = 1, when a = 1, it is the initial trajectory point; a = 1, 2, 3... A, where A represents the total number of trajectory points in the c-th aircraft taxiing trajectory point sequence;

[0021] Step 16: Obtain the trajectory point c of the c-th aircraft taxiing trajectory point sequence. a+1 ; Trajectory point c is filtered using road network topology information a Matching the trajectory point c in the road segment a+1 Candidate road segments with topological relationships are used to obtain a list of candidate road segments;

[0022] Based on the candidate road segment list, trajectory point c is obtained. a+1 Corresponding multiple projection distances;

[0023] Step 17: Combine the maximum error distance of the aircraft semantic trajectory points with the trajectory point c. a+1 After comparing, filtering, and sorting the corresponding multiple projected distances, the trajectory point c is obtained. a+1 The shortest path segment among the ranked candidate projection distances is represented by trajectory point c. a+1 The corresponding matching road segment;

[0024] Step 18: Verify trajectory point c using the constraints of the aircraft semantic trajectory representation. a+1 The correctness of the corresponding matching road segment result is checked. If correct, trajectory point c is obtained. a+1 If the corresponding road segment is correctly matched, proceed to step 19; otherwise, return to step 17 at trajectory point c. a+1 Remove incorrect trajectory points c from the sorted candidate projected distances. a+1 The corresponding matching road segment is used to obtain the updated trajectory point c. a+1 The sorted candidate projection distances are used to traverse the updated trajectory points c. a+1 The trajectory point c is obtained by sorting the candidate projection distances. a+1 The corresponding correct matching road segment;

[0025] Step 19: Determine the size of a+1 and A. If a+1 = A, end the calculation of the aircraft semantic trajectory representation, obtain the matching segments corresponding to multiple trajectory points in the c-th aircraft taxiing trajectory point sequence, generate the complete map matching result of the trajectory based on the matching segments corresponding to the multiple trajectory points, and perform deduplication processing on the complete map matching result of the trajectory to obtain the c-th initial aircraft semantic trajectory representation; if a+1 < A, let a = a+1, and repeat steps 16-19.

[0026] Step 20: Traverse multiple aircraft taxi trajectory data and repeat steps 12-19 to obtain multiple initial aircraft semantic track representations;

[0027] It is understood that the initial aircraft semantic track is represented by segment numbers to indicate the sequence of previous trajectory points. Furthermore, the projection distance expression is:

[0028]

[0029] Where D is the projection distance; x is the x-axis, y is the y-axis, and Q is the projection distance. x Let Q be the x-coordinate of the projection point Q of the trajectory point onto the candidate road segment. y Let d be the y-coordinate of the projection point Q of the trajectory point onto the candidate road segment. MAX This represents the maximum error distance of the aircraft's semantic trajectory points.

[0030] Preferably, the noise trajectory points are all trajectory points outside the airport area in the initial aircraft semantic trajectory representation; the map matching includes geometric matching algorithms and topological relationship algorithms;

[0031] The constraints on the aircraft semantic trajectory representation include: angle constraints on aircraft turning behavior, long-distance priority straight-line jump constraints, interpolation operations at complex curves, and jump point constraints on anti-interference road sections.

[0032] Furthermore, the angle constraint on the aircraft's turning behavior specifically includes: for trajectory point segment matching in complex road networks at curves, after adding connectivity verification through topological information during the trajectory point segment matching process, generating an angle vector based on endpoint coordinates, generating the cosine value of the vector angle through dot product operation, and obtaining the radian value of the angle using the inverse cosine function acosine, and using the radian value to correct the rationality of the aircraft's turning.

[0033] Furthermore, the expression for the angular constraint on the aircraft's turning behavior is:

[0034]

[0035] Where θ is the aircraft turning angle, vF represents the vector formed by the first and last coordinates of road segment F, and ||vF|| is the magnitude of the vector corresponding to road segment F; vB represents the vector formed by the first and last coordinates of road segment B, vF and vB are two connected road segments, and ||vB|| is the magnitude of the vector corresponding to road segment B.

[0036] The specific constraints of the long-distance priority straight-line jump segment for aircraft include: in response to the matching difficulties caused by the low sampling frequency of the field monitoring information, if the matching segments of adjacent sampling points are not connected during the trajectory point segment matching process, when it is determined that the disconnected segment is a straight taxiway, priority straight-line jump segment is performed and it is considered as a successful connection until matching continues;

[0037] The interpolation operation at complex curves of the aircraft specifically includes: to address the matching difficulties caused by the long sampling interval of the aircraft's surface taxiing trajectory at complex curves, when straight-line jump operations are not possible, interpolation operations are performed between the current trajectory point and the next trajectory point to increase the sampling point density and assist the matching work at complex curves;

[0038] The skip point constraint of the aircraft anti-interference section specifically includes: for possible abnormal sampling data, skipping unmatched trajectory points directly and continuing to match the next trajectory point; the unmatched trajectory points are abnormal trajectory point coordinate offsets that cause the projected distance to exceed the maximum error distance of the aircraft semantic trajectory point, causing the matching process to be interrupted;

[0039] If the matching process is interrupted, the time of the previous trajectory point matching result will be rolled back, the erroneous road segment will be marked, and the correct road will be searched again to correct the result of step 1. The number of consecutive rollbacks will not exceed 3.

[0040] Preferably, step 1, which involves cleaning and verifying the integrity of the initial aircraft semantic trajectory representation, specifically includes:

[0041] Obtain the initial semantic trajectory representation of the aircraft corresponding to the parking position, divide the parking position into parking position groups to obtain multiple parking position groups, and consider the pushback of aircraft in the same parking position to the same airport taxiway as the same behavior;

[0042] Establish key road network nodes for parking stand groups, obtain node segment numbers corresponding to key road network nodes for aircraft propulsion and pushback parking stand groups, delete node segment numbers corresponding to key road network nodes for aircraft propulsion and pushback parking stand groups, and realize semantic trajectory cleaning of parking stand areas.

[0043] Obtain the airport runway, select the key road segment at the end of the airport runway, and realize the integrity verification of the semantic trajectory representation of the aircraft corresponding to the end of the runway.

[0044] Step 2: Obtain the similarity matrix of the aircraft semantic trajectory representations using the multiple aircraft semantic trajectory representations; perform hierarchical clustering on the similarity matrix of the aircraft semantic trajectory representations to obtain multiple aircraft surface taxiing patterns;

[0045] Preferably, the specific steps for obtaining multiple aircraft surface taxiing modes include:

[0046] Step 21: Randomly select two aircraft semantic trajectory representations from multiple aircraft semantic trajectory representations to form a pairwise aircraft semantic trajectory sequence; thus obtaining multiple pairwise aircraft semantic trajectory sequences.

[0047] Preferably, the pairwise aircraft semantic trajectory sequences include an aircraft semantic trajectory sequence one and an aircraft semantic trajectory sequence two, which correspond to any two selected aircraft semantic trajectory representations.

[0048] The aircraft semantic trajectory sequence includes at least one candidate road segment, as well as the corresponding candidate road segment number and the corresponding candidate road segment coordinates;

[0049] The second semantic trajectory sequence of the aircraft includes at least one candidate road segment, as well as the corresponding candidate road segment number and the corresponding candidate road segment coordinates;

[0050] It is understood that the number of candidate road segments in the aircraft semantic trajectory sequence one and the aircraft semantic trajectory sequence two may be equal or unequal;

[0051] Step 22: Select the e-th pair of aircraft semantic trajectory sequences and calculate based on the improved longest common subsequence (LCS) to obtain the total length of the common road segment corresponding to the e-th pair of aircraft semantic trajectory sequences, and obtain the similarity of the corresponding total length of the common road segment, e = 1, 2, 3...E, where E represents the total number of pair of aircraft semantic trajectory sequences;

[0052] Step 23: Traverse multiple pairs of aircraft semantic trajectory sequences, repeat steps 21-22, and obtain the total length of the common road segment corresponding to the multiple pairs of aircraft semantic trajectory sequences. Based on the total length of the common road segment corresponding to the multiple pairs of aircraft semantic trajectory sequences, generate a similarity matrix of all pairs of aircraft semantic trajectory representations.

[0053] Step 24: Perform hierarchical clustering on the similarity matrices of all pairwise aviation semantic trajectory representations to obtain multiple aircraft surface taxiing patterns.

[0054] Furthermore, the total length of the common road segment corresponding to the pairwise semantic trajectory sequences of the aircraft mentioned in step 22 is expressed as:

[0055]

[0056] in, The total length of the common road segment corresponding to each pair of aviation semantic trajectory sequences. Let be the segment length corresponding to segment i in the m-th aircraft semantic trajectory representation, that is, the length of the segment up to the m-th aircraft semantic trajectory representation in the pairwise aircraft semantic trajectory sequence. Let m be the segment length corresponding to segment j in the semantic trajectory representation of the o-th aircraft, that is, the length of the segment up to the o-th aircraft semantic trajectory representation in the pairwise aircraft semantic trajectory sequence. i Let i be the candidate segment number in the m-th aircraft semantic trajectory representation, o j Let j be the candidate segment number in the semantic trajectory representation of the o-th aircraft, i = 0, 1, 2, ..., I, where I is the total number of candidate segments in the semantic trajectory representation of the m-th aircraft, j = 0, 1, 2, ..., J, where J is the total number of candidate segments in the semantic trajectory representation of the o-th aircraft. Let L be the total length of the common road segment corresponding to the road segment number i-1 represented by the m-th aircraft semantic trajectory and the road segment number j-1 represented by the o-th aircraft semantic trajectory. ij This represents the length of the common segment corresponding to segment i number represented by the m-th aircraft semantic trajectory and segment j number represented by the o-th aircraft semantic trajectory; Let be the total length of the common road segment corresponding to the road segment i number represented by the m-th aircraft semantic trajectory and the road segment j-1 number represented by the o-th aircraft semantic trajectory, and the longest common road segment among the total lengths of the common road segments corresponding to the road segment i-1 number represented by the m-th aircraft semantic trajectory and the road segment j number represented by the o-th aircraft semantic trajectory.

[0057] It is understandable that when the candidate segment i number and candidate segment j number are the same in the semantic trajectory sequences of two aircraft, that is, m i =o j At that time, based on the length of the common road segment corresponding to the previous iteration step For the length L of the common road segment corresponding to the current iteration step ij Update the current iteration and obtain the total length of the common road segment corresponding to the current iteration step. If i,j>0,m i ≠o j If the candidate segment i and candidate segment j in the semantic trajectory sequences of each pair of aircraft are different, it means that there are no common segments in the current iteration step. The longest total length of the common segments in the previous iteration step is selected. The final total length of the common road segment in the current iteration step

[0058] The similarity expression for the semantic trajectories of each pair of aircraft is as follows:

[0059]

[0060] Among them, M(T) m ,T o M(T) represents the total length of the common segment between the m-th aircraft semantic trajectory representation and the o-th aircraft semantic trajectory representation. m ,T m M(T) represents the total length of the common segment between the m-th aircraft semantic trajectory representation and itself, i.e., the total length of the m-th aircraft semantic trajectory representation. o ,T o Let be the total length of the common road segment between the o-th aircraft semantic trajectory representation and itself, that is, the total length of the o-th aircraft semantic trajectory representation; take the total length of the common road segment between each pair of aircraft semantic trajectory representations as the upper half of the matrix, the diagonal of the matrix is ​​the length of the corresponding aircraft semantic trajectory representation itself, and the lower half of the matrix is ​​the similarity between each pair of aircraft semantic trajectory representations.

[0061] Furthermore, the similarity matrices of all pairwise aviation semantic trajectory representations described in step 24 are subjected to hierarchical clustering to obtain the surface taxiing patterns of multiple aircraft. Specific steps include:

[0062] Each aircraft semantic trajectory is represented as a single cluster, and each pair of aircraft semantic trajectories is represented as a pair of clusters.

[0063] Based on the similarity matrix of the semantic trajectory representation of each pair of aircraft, the distance between the corresponding pairs of clusters is obtained. The pairs of clusters with the closest distance (i.e. the highest similarity) are then clustered to obtain the clustering results.

[0064] The clustering results are represented as the surface taxiing patterns corresponding to the semantic trajectories of aircraft.

[0065] It is understood that the aircraft's surface taxiing mode is the taxiing mode in which the aircraft pushes out from the parking position to the runway for takeoff or taxis from the runway to the parking position after landing.

[0066] Step 3: Construct semantic trajectory outlier values ​​based on the similarity matrix of the aircraft semantic trajectory representation described in Step 2 and multiple aircraft taxiing patterns. Use the semantic trajectory outlier values ​​to perform anomaly detection on the taxiing patterns corresponding to the semantic trajectory representation of the aircraft to be detected.

[0067] It is understood that the semantic trajectory of the aircraft to be detected is represented as the semantic trajectory of the aircraft that did not participate in the calculation in step 2;

[0068] The specific steps include:

[0069] Step 31: Calculate the similarity matrix based on the pairwise aircraft semantic trajectory representations described in Step 23.

[0070] The maximum reachable outlier value corresponding to each aircraft surface taxiing mode described in step 24 is obtained; the maximum reachable outlier value is used as the judgment criterion for anomaly detection of aircraft surface taxiing modes;

[0071] Step 32: Select the b-th aircraft semantic trajectory representation to be detected, and obtain the average similarity between the b-th aircraft semantic trajectory representation and other surface taxiing modes to obtain multiple average similarities. Arrange the multiple average similarities in order of magnitude to obtain the maximum average similarity. Calculate the outlier value corresponding to the maximum average similarity, and compare the corresponding outlier value with the maximum reachable outlier value in step 31 to determine whether the surface taxiing mode corresponding to the b-th aircraft semantic trajectory representation is abnormal, thereby realizing taxiing mode anomaly detection.

[0072] The judgment criterion for abnormal detection of the gliding mode is expressed as follows:

[0073]

[0074] in, Let b be the average similarity between the semantic trajectory representation of the b-th aircraft to be detected and other scene taxiing patterns, where b = 1, 2, 3…B, and B represents the total number of semantic trajectory representations of the aircraft to be detected. For the semantic trajectory representation of the b-th aircraft to be detected, O Max This represents the maximum reachable outlier.

[0075] Compared with the prior art, the present invention has at least the following beneficial effects:

[0076] (1) This invention avoids problems such as data loss and offset noise by associating the original field monitoring data with the field map data, and uses topological information to ensure the continuity and integrity of the data. In addition, the semantic trajectory is more regular in form, making it easier to calculate the similarity relationship between trajectories, which is beneficial for analyzing and identifying abnormal sliding patterns;

[0077] (2) The pattern recognition results of the present invention are the basis for abnormal pattern detection. In the process, a hierarchical clustering method is used after establishing a track similarity matrix. The pattern recognition accuracy can be adjusted according to the needs to achieve adaptive scene taxiing pattern recognition.

[0078] (3) Taking into account the different pattern distributions and definitions of anomalous patterns at different airports, this invention calculates outlier values ​​by means of the average similarity within each pattern, and takes the largest outlier value as the detection standard for taxiing pattern anomalies. By using a data-driven approach, unique and adaptive anomaly detection parameters are established for each airport environment, ensuring the breadth, reliability, and real-time nature of the results. Attached Figure Description

[0079] The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of the invention.

[0080] Figure 1 This is a schematic diagram of the aircraft semantic trajectory representation and taxiing mode anomaly detection method of the present invention;

[0081] Figure 2 This is a flowchart of the aircraft semantic trajectory representation and taxiing mode anomaly detection method in an embodiment of the present invention. Detailed Implementation

[0082] To better understand the above-described objectives, features, and advantages of the present invention, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments of the present invention and the features thereof can be combined with each other. Furthermore, the present invention can be implemented in other ways different from those described herein; therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0083] A specific embodiment of the present invention, such as Figure 1-2 This invention discloses a method for semantic trajectory representation and taxiing pattern anomaly detection on airport surfaces. To illustrate the effectiveness of the proposed method, a specific embodiment is provided below for detailed explanation of the above technical solution. The specific implementation steps are as follows:

[0084] Step 1: Obtain multiple initial taxiing trajectories of aircraft based on the surface surveillance radar system; obtain multiple aircraft taxiing trajectory data by matching the multiple initial taxiing trajectories of aircraft through map matching;

[0085] By utilizing multiple aircraft taxi trajectory data, multiple initial aircraft semantic trajectory representations are obtained;

[0086] Based on the division of parking positions, multiple initial aircraft semantic trajectory representations are cleaned and their integrity is verified to obtain multiple aircraft semantic trajectory representations;

[0087] Preferably, the specific steps for obtaining multiple initial aircraft semantic trajectory representations include:

[0088] Step 11: Under the constraints of aircraft semantic trajectory representation, obtain multiple initial taxiing trajectories of the aircraft based on the surface surveillance radar system;

[0089] It is understandable that the multiple aircraft taxiing trajectories are derived from takeoff / landing behavior;

[0090] Airport contour coordinates are established by map matching, and noisy trajectory points in multiple aircraft initial taxi trajectories are deleted using the airport contour coordinates to obtain multiple aircraft taxi trajectory data.

[0091] Based on the multiple aircraft taxiing trajectory data, a sequence of multiple aircraft taxiing trajectory points is generated;

[0092] Step 12: Establish a candidate road segment set by taking all road segments of the road network as candidate road segments; the first trajectory point in each aircraft taxiing trajectory point sequence starts from the runway end; each aircraft taxiing trajectory point sequence includes multiple runway ends and parking position ends;

[0093] Step 13: Select the c-th aircraft taxiing trajectory point sequence, where c = 1, 2, 3…C, and C represents the total number of aircraft taxiing trajectory point sequences;

[0094] Obtain the initial trajectory point c1 of the c-th aircraft taxiing trajectory point sequence, and obtain the minimum projection distance from the initial trajectory point c1 to each candidate road segment in the candidate road segment set using the Euclidean distance method, thus obtaining multiple projection distances corresponding to the initial trajectory point c1.

[0095] Step 14: Set the maximum error distance d for the aircraft semantic trajectory points. max The maximum error distance of the aircraft semantic trajectory point is compared and filtered with multiple projection distances corresponding to the initial trajectory point c1 to select multiple candidate projection distances for the initial trajectory point c1.

[0096] Sort the multiple candidate projection distances of the initial trajectory point c1 to obtain the sorted candidate projection distances; the smallest candidate projection distance among the sorted candidate projection distances is represented as the matching road segment of the initial trajectory point c1.

[0097] Furthermore, the multiple candidate projection distances of the initial trajectory point c1 are the maximum error distance d among the multiple projection distances corresponding to the initial trajectory point c1 that is less than the aircraft semantic trajectory point. max When, the candidate projection distance of the corresponding initial trajectory point.

[0098] Step 15: Let a = 1, when a = 1, it is the initial trajectory point; a = 1, 2, 3... A, where A represents the total number of trajectory points in the c-th aircraft taxiing trajectory point sequence;

[0099] Step 16: Obtain the trajectory point c of the c-th aircraft taxiing trajectory point sequence. a+1 ; Trajectory point c is filtered using road network topology information a Matching the trajectory point c in the road segment a+1 Candidate road segments with topological relationships are used to obtain a list of candidate road segments;

[0100] Based on the candidate road segment list, trajectory point c is obtained. a+1 Corresponding multiple projection distances;

[0101] Step 17: Combine the maximum error distance of the aircraft semantic trajectory points with the trajectory point c. a+1 After comparing, filtering, and sorting the corresponding multiple projected distances, the trajectory point c is obtained. a+1 The shortest path segment among the ranked candidate projection distances is represented by trajectory point c. a+1 The corresponding matching road segment;

[0102] Step 18: Verify trajectory point c using the constraints of the aircraft semantic trajectory representation. a+1 The correctness of the corresponding matching road segment result is checked. If correct, trajectory point c is obtained. a+1 If the corresponding road segment is correctly matched, proceed to step 19; otherwise, return to step 17 at trajectory point c. a+1 Remove incorrect trajectory points c from the sorted candidate projected distances. a+1 The corresponding matching road segment is used to obtain the updated trajectory point c. a+1 The sorted candidate projection distances are used to traverse the updated trajectory points c. a+1 The trajectory point c is obtained by sorting the candidate projection distances. a+1 The corresponding correct matching road segment;

[0103] Step 19: Determine the size of a+1 and A. If a+1 = A, end the calculation of the aircraft semantic trajectory representation, obtain the matching segments corresponding to multiple trajectory points in the c-th aircraft taxiing trajectory point sequence, generate the complete map matching result of the trajectory based on the matching segments corresponding to the multiple trajectory points, and perform deduplication processing on the complete map matching result of the trajectory to obtain the c-th initial aircraft semantic trajectory representation; if a+1 < A, let a = a+1, and repeat steps 16-19.

[0104] Step 20: Traverse multiple aircraft taxi trajectory data and repeat steps 12-19 to obtain multiple initial aircraft semantic track representations;

[0105] It is understood that the initial aircraft semantic track is represented by using segment numbers to represent the previous sequence of trajectory points.

[0106] Furthermore, the projection distance expression is:

[0107]

[0108] Where D is the projection distance; x is the x-axis, y is the y-axis, and Q is the projection distance. x Let Q be the x-coordinate of the projection point Q of the trajectory point onto the candidate road segment. y Let d be the y-coordinate of the projection point Q of the trajectory point onto the candidate road segment. MAX This represents the maximum error distance of the aircraft's semantic trajectory points.

[0109] Preferably, the noise trajectory points are all trajectory points outside the airport area in the initial aircraft semantic trajectory representation; the map matching includes geometric matching algorithms and topological relationship algorithms;

[0110] The constraints on the aircraft semantic trajectory representation include: angle constraints on aircraft turning behavior, long-distance priority straight-line jump constraints, interpolation operations at complex curves, and jump point constraints on anti-interference road sections.

[0111] Furthermore, the angle constraint on the aircraft's turning behavior specifically includes: for trajectory point segment matching in complex road networks at curves, after adding connectivity verification through topological information during the trajectory point segment matching process, generating an angle vector based on endpoint coordinates, generating the cosine value of the vector angle through dot product operation, and obtaining the radian value of the angle using the inverse cosine function acosine, and using the radian value to correct the rationality of the aircraft's turning.

[0112] Furthermore, the expression for the angular constraint on the aircraft's turning behavior is:

[0113]

[0114] Where θ is the aircraft turning angle, vF represents the vector formed by the first and last coordinates of road segment F, and ||vF|| is the magnitude of the vector corresponding to road segment F; vB represents the vector formed by the first and last coordinates of road segment B, vF and vB are two connected road segments, and ||vB|| is the magnitude of the vector corresponding to road segment B.

[0115] The specific constraints of the long-distance priority straight-line jump segment for aircraft include: in response to the matching difficulties caused by the low sampling frequency of the field monitoring information, if the matching segments of adjacent sampling points are not connected during the trajectory point segment matching process, when it is determined that the disconnected segment is a straight taxiway, priority straight-line jump segment is performed and it is considered as a successful connection until matching continues;

[0116] The interpolation operation at complex curves of the aircraft specifically includes: to address the matching difficulties caused by the long sampling interval of the aircraft's surface taxiing trajectory at complex curves, when straight-line jump operations are not possible, interpolation operations are performed between the current trajectory point and the next trajectory point to increase the sampling point density and assist the matching work at complex curves;

[0117] The skip point constraint of the aircraft anti-interference section specifically includes: for possible abnormal sampling data, skipping unmatched trajectory points directly and continuing to match the next trajectory point; the unmatched trajectory points are abnormal trajectory point coordinate offsets that cause the projected distance to exceed the maximum error distance of the aircraft semantic trajectory point, causing the matching process to be interrupted;

[0118] If the matching process is interrupted, the time of the previous trajectory point matching result will be rolled back, the erroneous road segment will be marked, and the correct road will be searched again to correct the result of step 1. The number of consecutive rollbacks will not exceed 3.

[0119] Preferably, step 1, which involves cleaning and verifying the integrity of the initial aircraft semantic trajectory representation, specifically includes:

[0120] Obtain the initial semantic trajectory representation of the aircraft corresponding to the parking position, divide the parking position into parking position groups to obtain multiple parking position groups, and consider the pushback of aircraft in the same parking position to the same airport taxiway as the same behavior;

[0121] Establish key road network nodes for parking stand groups, obtain node segment numbers corresponding to key road network nodes for aircraft propulsion and pushback parking stand groups, delete node segment numbers corresponding to key road network nodes for aircraft propulsion and pushback parking stand groups, and realize semantic trajectory cleaning of parking stand areas.

[0122] Obtain the airport runway, select the key road segment at the end of the airport runway, and realize the integrity verification of the semantic trajectory representation of the aircraft corresponding to the end of the runway.

[0123] Step 2: Obtain the similarity matrix of the aircraft semantic trajectory representations using the multiple aircraft semantic trajectory representations; perform hierarchical clustering on the similarity matrix of the aircraft semantic trajectory representations to obtain multiple aircraft surface taxiing patterns;

[0124] Preferably, the specific steps for obtaining multiple aircraft surface taxiing modes include:

[0125] Step 21: Randomly select two aircraft semantic trajectory representations from multiple aircraft semantic trajectory representations to form a pairwise aircraft semantic trajectory sequence; thus obtaining multiple pairwise aircraft semantic trajectory sequences.

[0126] Preferably, the pairwise aircraft semantic trajectory sequences include an aircraft semantic trajectory sequence one and an aircraft semantic trajectory sequence two, which correspond to any two selected aircraft semantic trajectory representations.

[0127] The aircraft semantic trajectory sequence includes at least one candidate road segment, as well as the corresponding candidate road segment number and the corresponding candidate road segment coordinates;

[0128] The second semantic trajectory sequence of the aircraft includes at least one candidate road segment, as well as the corresponding candidate road segment number and the corresponding candidate road segment coordinates;

[0129] It is understood that the number of candidate road segments in the aircraft semantic trajectory sequence one and the aircraft semantic trajectory sequence two may be equal or unequal;

[0130] Step 22: Select the e-th pair of aircraft semantic trajectory sequences and calculate based on the improved longest common subsequence (LCS) to obtain the total length of the common road segment corresponding to the e-th pair of aircraft semantic trajectory sequences, and obtain the similarity of the corresponding total length of the common road segment, e = 1, 2, 3...E, where E represents the total number of pair of aircraft semantic trajectory sequences;

[0131] Step 23: Traverse multiple pairs of aircraft semantic trajectory sequences, repeat steps 21-22, and obtain the total length of the common road segment corresponding to the multiple pairs of aircraft semantic trajectory sequences. Based on the total length of the common road segment corresponding to the multiple pairs of aircraft semantic trajectory sequences, generate a similarity matrix of all pairs of aircraft semantic trajectory representations.

[0132] Step 24: Perform hierarchical clustering on the similarity matrices of all pairwise aviation semantic trajectory representations to obtain multiple aircraft surface taxiing patterns.

[0133] Furthermore, the total length of the common road segment corresponding to the pairwise semantic trajectory sequences of the aircraft mentioned in step 22 is expressed as:

[0134]

[0135] in, The total length of the common road segment corresponding to each pair of aviation semantic trajectory sequences. Let be the segment length corresponding to segment i in the m-th aircraft semantic trajectory representation, that is, the length of the segment up to the m-th aircraft semantic trajectory representation in the pairwise aircraft semantic trajectory sequence. Let be the segment length corresponding to segment j in the semantic trajectory representation of the o-th aircraft, that is, the length of the segment up to the o-th aircraft semantic trajectory representation in the pairwise aviation semantic trajectory sequence. i Let i be the candidate segment number in the m-th aircraft semantic trajectory representation, o j Let j be the candidate segment number in the semantic trajectory representation of the o-th aircraft, i = 0, 1, 2, ..., I, where I is the total number of candidate segments in the semantic trajectory representation of the m-th aircraft, j = 0, 1, 2, ..., J, where J is the total number of candidate segments in the semantic trajectory representation of the o-th aircraft. Let L be the total length of the common road segment corresponding to the road segment number i-1 represented by the m-th aircraft semantic trajectory and the road segment number j-1 represented by the o-th aircraft semantic trajectory. ij This represents the length of the common segment corresponding to segment i number represented by the m-th aircraft semantic trajectory and segment j number represented by the o-th aircraft semantic trajectory; Let be the total length of the common road segment corresponding to the road segment i number represented by the m-th aircraft semantic trajectory and the road segment j-1 number represented by the o-th aircraft semantic trajectory, and the longest common road segment among the total lengths of the common road segments corresponding to the road segment i-1 number represented by the m-th aircraft semantic trajectory and the road segment j number represented by the o-th aircraft semantic trajectory.

[0136] It is understandable that when the candidate segment i number and candidate segment j number are the same in the semantic trajectory sequences of two aircraft, that is, m i =o j At that time, based on the length of the common road segment corresponding to the previous iteration step For the length L of the common road segment corresponding to the current iteration step ij Update the current iteration and obtain the total length of the common road segment corresponding to the current iteration step. If i,j>0,m i ≠o j If the candidate segment i and candidate segment j in the semantic trajectory sequences of each pair of aircraft are different, it means that there are no common segments in the current iteration step. The longest total length of the common segments in the previous iteration step is selected. The final total length of the common road segment in the current iteration step

[0137] The similarity expression for the semantic trajectories of each pair of aircraft is as follows:

[0138]

[0139] Among them, M(T) m ,T o M(T) represents the total length of the common segment between the m-th aircraft semantic trajectory representation and the o-th aircraft semantic trajectory representation. m ,T m M(T) represents the total length of the common segment between the m-th aircraft semantic trajectory representation and itself, i.e., the total length of the m-th aircraft semantic trajectory representation. o ,T o Let be the total length of the common road segment between the o-th aircraft semantic trajectory representation and itself, that is, the total length of the o-th aircraft semantic trajectory representation; take the total length of the common road segment between each pair of aircraft semantic trajectory representations as the upper half of the matrix, the diagonal of the matrix is ​​the length of the corresponding aircraft semantic trajectory representation itself, and the lower half of the matrix is ​​the similarity between each pair of aircraft semantic trajectory representations.

[0140] Furthermore, the similarity matrices of all pairwise aviation semantic trajectory representations described in step 24 are subjected to hierarchical clustering to obtain the surface taxiing patterns of multiple aircraft. Specific steps include:

[0141] Each aircraft semantic trajectory is represented as a single cluster, and each pair of aircraft semantic trajectories is represented as a pair of clusters.

[0142] Based on the similarity matrix of the semantic trajectory representation of each pair of aircraft, the distance between the corresponding pairs of clusters is obtained. The pairs of clusters with the closest distance (i.e. the highest similarity) are then clustered to obtain the clustering results.

[0143] The clustering results are represented as the surface taxiing patterns corresponding to the semantic trajectories of aircraft.

[0144] It is understood that the aircraft's surface taxiing mode is the taxiing mode in which the aircraft pushes out from the parking position to the runway for takeoff or taxis from the runway to the parking position after landing.

[0145] Step 3: Construct semantic trajectory outlier values ​​based on the similarity matrix of the aircraft semantic trajectory representation described in Step 2 and multiple aircraft taxiing patterns. Use the semantic trajectory outlier values ​​to perform anomaly detection on the taxiing patterns corresponding to the semantic trajectory representation of the aircraft to be detected.

[0146] It is understood that the semantic trajectory of the aircraft to be detected is represented as the semantic trajectory of the aircraft that did not participate in the calculation in step 2;

[0147] The specific steps include:

[0148] Step 31: Calculate the similarity matrix based on the pairwise aircraft semantic trajectory representations described in Step 23.

[0149] The maximum reachable outlier value corresponding to each aircraft surface taxiing mode described in step 24 is obtained; the maximum reachable outlier value is used as the judgment criterion for anomaly detection of aircraft surface taxiing modes;

[0150] Step 32: Select the b-th aircraft semantic trajectory representation to be detected, and obtain the average similarity between the b-th aircraft semantic trajectory representation and other surface taxiing modes to obtain multiple average similarities. Arrange the multiple average similarities in order of magnitude to obtain the maximum average similarity. Calculate the outlier value corresponding to the maximum average similarity, and compare the corresponding outlier value with the maximum reachable outlier value in step 31 to determine whether the surface taxiing mode corresponding to the b-th aircraft semantic trajectory representation is abnormal, thereby realizing taxiing mode anomaly detection.

[0151] The judgment criterion for abnormal detection of the gliding mode is expressed as follows:

[0152]

[0153] in, Let b be the average similarity between the semantic trajectory representation of the b-th aircraft to be detected and other scene taxiing patterns, where b = 1, 2, 3…B, and B represents the total number of semantic trajectory representations of the aircraft to be detected. For the semantic trajectory representation of the b-th aircraft to be detected, O Max This represents the maximum reachable outlier.

[0154] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for semantic trajectory representation and taxiing pattern anomaly detection on airport surfaces, characterized in that, include: Step 1: Based on the surface surveillance radar system and map matching, obtain multiple initial aircraft semantic trajectory representations; based on the parking stand group division, clean and verify the integrity of the multiple initial aircraft semantic trajectory representations to obtain multiple aircraft semantic trajectory representations. Step 2: Obtain the similarity matrix of the aircraft semantic trajectory representations using the multiple aircraft semantic trajectory representations; perform hierarchical clustering on the similarity matrix of the aircraft semantic trajectory representations to obtain multiple aircraft surface taxiing patterns; Step 3: Construct semantic trajectory outlier values ​​based on the similarity matrix of the aircraft semantic trajectory representation described in Step 2 and multiple aircraft taxiing patterns. Use the semantic trajectory outlier values ​​to perform anomaly detection on the taxiing patterns corresponding to the semantic trajectory representation of the aircraft to be detected. Step 2, which describes obtaining multiple aircraft surface taxiing patterns, includes the following specific steps: Step 21: Randomly select two aircraft semantic trajectory representations from multiple aircraft semantic trajectory representations to form a pairwise aircraft semantic trajectory sequence; thus obtaining multiple pairwise aircraft semantic trajectory sequences. Step 22, Select the first e The semantic trajectory sequences of each pair of aircraft are calculated based on the improved longest common subsequence (LCS) to obtain the first... e The similarity of the total length of the common road segments corresponding to each pair of aircraft semantic trajectory sequences is obtained by calculating the total length of the corresponding common road segments. e =1,2,3… E , E This represents the total number of semantic trajectory sequences for each pair of aircraft. Step 23: Traverse multiple pairs of aircraft semantic trajectory sequences, repeat steps 21-22, and obtain the total length of the common road segment corresponding to the multiple pairs of aircraft semantic trajectory sequences. Based on the total length of the common road segment corresponding to the multiple pairs of aircraft semantic trajectory sequences, generate a similarity matrix of all pairs of aircraft semantic trajectory representations. Step 24: Perform hierarchical clustering on the similarity matrices of all pairwise aviation semantic trajectory representations to obtain multiple aircraft surface taxiing patterns; The total length of the common segment corresponding to the semantic trajectory sequences of each pair of aircraft is expressed as follows: in, The total length of the common road segment corresponding to each pair of aviation semantic trajectory sequences. For the first m The segment represented by the semantic trajectory of an aircraft i The length of the road segment corresponding to the number For the first o The segment represented by the semantic trajectory of an aircraft j The length of the road segment corresponding to the number i For the first m Candidate segments in the semantic trajectory representation of an aircraft i serial number, For the first o Candidate segments in the semantic trajectory representation of an aircraft j serial number, i =0,1,2,… I , I For the first m The total number of candidate road segments in the semantic trajectory representation of an aircraft. j =0,1,2,… J , J For the first o The total number of candidate road segments in the semantic trajectory representation of an aircraft. For the first m The segment represented by the semantic trajectory of an aircraft i -1 number and the o The segment represented by the semantic trajectory of an aircraft j The total length of the public road segment corresponding to number -1 Indicates the first m The segment represented by the semantic trajectory of an aircraft i Number and o The segment represented by the semantic trajectory of an aircraft j The length of the public road segment corresponding to the number; For the first m The segment represented by the semantic trajectory of an aircraft i Number and o The segment represented by the semantic trajectory of an aircraft j The total length of the public road segment corresponding to number -1, the th m The segment represented by the semantic trajectory of an aircraft i -1 number and the o The segment represented by the semantic trajectory of an aircraft j The total length of the longest public road segment corresponding to the number; The similarity expression for the total length of the corresponding public road segments is: in, For the first m Semantic trajectory representation of aircraft and the first o The total length of the common road segment represented by the semantic trajectory of an aircraft. For the first m The total length of the semantic trajectory representation of an aircraft. For the first o The total length of the semantic trajectory representation of an aircraft; The specific steps for anomaly detection include: Step 31: Calculate the similarity matrix based on the pairwise aircraft semantic trajectory representations described in Step 23. Obtain the maximum reachable outlier value corresponding to each aircraft surface taxiing mode described in step 24; Step 32: Select the semantic trajectory representation of the aircraft to be detected, and obtain the average similarity of the semantic trajectory representation of the aircraft to be detected with other taxiing modes. Multiple average similarities are obtained. The multiple average similarities are arranged in order of size to obtain the maximum average similarity. The outlier value corresponding to the maximum average similarity is calculated. The corresponding outlier value is compared with the maximum reachable outlier value in step 31 to determine whether the taxiing mode corresponding to the semantic trajectory representation of the aircraft to be detected is abnormal, thereby realizing taxiing mode anomaly detection.

2. The detection method according to claim 1, characterized in that, The specific steps for obtaining multiple initial aircraft semantic track representations include: Step 11: Establish semantic trajectory representation constraints for aircraft, and obtain multiple aircraft taxiing trajectory point sequences based on the surface surveillance radar system and airport map outline coordinates; Step 12: Select all road segments in the road network as candidate road segments and establish a candidate road segment set; Step 13, obtain the first c Initial trajectory points of the aircraft taxiing trajectory point sequence c 1, c =1,2,3… C , C This represents the total number of aircraft taxiing trajectory point sequences; the initial trajectory points are obtained based on the Euclidean distance method. c 1. The multiple projected distances corresponding to each candidate road segment in the candidate road segment set; Step 14: Set the maximum error distance of the aircraft semantic trajectory points; based on the maximum error distance of the aircraft semantic trajectory points and the initial trajectory points... c The initial trajectory point is obtained by corresponding to multiple projection distances. c 1. Matching road segments; Step 15, Order a =1, when a =1 is the initial trajectory point; Step 16, obtain the first c Trajectory points of a sequence of aircraft taxiing trajectory points c a+1 ; Filtering trajectory points using road network topology information c a Matching road segments with trajectory points c a+1 Candidate road segments with topological relationships are used to obtain trajectory points. c a+1 Corresponding multiple projection distances; Step 17: Based on the maximum error distance of the aircraft semantic trajectory points and the trajectory points c a+1 Trajectory points are obtained from multiple corresponding projection distances. c a+1 The corresponding matching road segment; Step 18: Use the aircraft semantic trajectory representation to represent the constraints on the trajectory points. c a+1 The corresponding matching road segments are verified to obtain trajectory points. c a+1 The corresponding correct matching road segment; Step 19, Judgment a +1 and A Size, A Represents the sequence of aircraft taxiing trajectory points c The total number of trajectory points; if a +1= A Then the calculation of the aircraft semantic trajectory representation ends, and the first... c The initial semantic track representation of the aircraft corresponding to the sequence of aircraft taxiing trajectory points; if a +1 < A ,make a = a +1, repeat steps 16-19; Step 20: Traverse multiple aircraft taxiing trajectory point sequences and repeat steps 12-19 to obtain multiple initial aircraft semantic trajectory representations.

3. The detection method according to claim 2, characterized in that, The map matching is a geometric matching algorithm or a topological relationship algorithm.

4. The detection method according to claim 1, characterized in that, The constraints on the semantic trajectory representation of the aircraft include: angular constraints on the aircraft's turning behavior, constraints on the aircraft's long-distance priority straight-line jump segments, interpolation operations at complex curves and / or jump point constraints on the aircraft's anti-interference road segments.

5. The detection method according to claim 4, characterized in that, The expression for the angular constraint on the aircraft's turning behavior is: in, For the aircraft's turning angle, For road section F Using the vector formed by the coordinates of the first and last ends, For road section F The magnitude of the corresponding vector; For road section B Using the vector formed by the coordinates of the first and last ends, and These are two connected road sections. For road section B The magnitude of the corresponding vector.

Images