CN116244391A - Method for extracting typical array position of massive track targets - Google Patents
Method for extracting typical array position of massive track targets Download PDFInfo
- Publication number
- CN116244391A CN116244391A CN202211623673.3A CN202211623673A CN116244391A CN 116244391 A CN116244391 A CN 116244391A CN 202211623673 A CN202211623673 A CN 202211623673A CN 116244391 A CN116244391 A CN 116244391A
- Authority
- CN
- China
- Prior art keywords
- track
- closed
- compressed
- point
- cluster
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000000605 extraction Methods 0.000 claims abstract description 23
- 238000009499 grossing Methods 0.000 claims abstract description 18
- 238000012545 processing Methods 0.000 claims abstract description 18
- 230000006835 compression Effects 0.000 claims abstract description 13
- 238000007906 compression Methods 0.000 claims abstract description 13
- 238000013507 mapping Methods 0.000 claims abstract description 5
- 239000011159 matrix material Substances 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 12
- 230000002159 abnormal effect Effects 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 abstract description 18
- 230000000694 effects Effects 0.000 abstract description 6
- 238000005065 mining Methods 0.000 abstract description 5
- 238000004458 analytical method Methods 0.000 description 10
- 238000004422 calculation algorithm Methods 0.000 description 7
- UHPUCGUHOAGWID-UHFFFAOYSA-N mbid Chemical compound O.OC.C12C34CCN2CC=CC1(CC)C(OC(C)=O)C(C(=O)OC)(O)C4N(C)C(C=C1OC)=C3C=C1C12C3=CC=CC=C3NC11C(C(=O)OC)CC(CC(O)(CO)CC)CN1CC2 UHPUCGUHOAGWID-UHFFFAOYSA-N 0.000 description 4
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007621 cluster analysis Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000013432 robust analysis Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/29—Geographical information databases
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
- H03M7/00—Conversion of a code where information is represented by a given sequence or number of digits to a code where the same, similar or subset of information is represented by a different sequence or number of digits
- H03M7/30—Compression; Expansion; Suppression of unnecessary data, e.g. redundancy reduction
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Remote Sensing (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Train Traffic Observation, Control, And Security (AREA)
Abstract
The invention provides a method for extracting typical array positions of a mass track target, which comprises the following steps: step 1, reading historical track point data from a historical track database; step 2, mapping the read track point data into track data according to the track ID; step 3, for any track, track smoothing, track compression and closed track extraction are sequentially carried out to obtain a compressed track and a closed track in the compressed track; step 4, the tracks are summarized by taking the year and the object ID as the combined main key, and a compressed track set and a closed track set of each object in the corresponding year are obtained; and 5, processing the compressed track set and the closed route set by adopting a clustering and array position combining mode aiming at the target of any year to obtain all array position areas and frequent active routes in the areas of the target. The method provided by the invention has the advantages of high recognition rate of typical array positions, good robustness of the calculation method and high mining speed, and can be used for analyzing and assisting in recognition of the activity rule of the moving target.
Description
Technical Field
The invention relates to the field of intelligent analysis and big data mining, in particular to a method for extracting typical array positions of a massive track target.
Background
The typical array position is a frequent active area of a classical target in the track analysis field, the activity of the classical target is regular every trip, the motion trail is stable, the frequent active area of the classical target is the typical array position, the typical array position plays a very important role in the analysis of target identification, target warning, target behavior intention and the like, and the typical array position extraction based on the historical track is to extract the typical array position of the frequent activity of the target by excavating in a large number of historical activity tracks of the classical target.
There are a number of difficulties in analyzing historical track data of classical targets, mainly in:
1) The historical track data volume is huge. The historical track is typically stored in the database in the form of track points. When the typical array position of the target is mined, the historical tracks of the target are generally summarized by year, at the moment, the number of the historical track points participating in calculation is huge, and the time efficiency of analysis is low.
2) The original track data has errors. The track data is received by multiple sensors, and the original track is usually not an ideal smooth curve due to the failure of the sensors or the existence of measurement errors, and some obvious abnormal track points or zigzag track segments exist. Direct statistical analysis of raw track data that has not been processed can lead to deviations or even errors in the array extraction results.
In addition, typical array position extraction of the track has the following differences from traditional ground track interest point mining:
1) There is no obvious standing point in the airplane type target track. The point of interest mining of the ground track usually takes the position where the target object stays for a long time in the track as a standing point, however, the airplane target object is rarely kept in a similar position continuously, and the standing point cannot be formed due to the fact that the target object stays for a long time in the point of interest, so that a larger region of interest needs to be mined instead of a simple position standing point.
2) The track cannot be map matched by means of road network information. The flight of the airplane targets is independent of the road network, the airplane targets cannot move along the road, and the track data of the airplane targets are not constrained by the road network. Topology similarity exists among different tracks, but key point extraction and matching cannot be performed according to road network information.
3) Aircraft targets often orbit around a point of interest location. This phenomenon results from the specific purpose or intent of the relevant target, such as the reconnaissance class aircraft performing loop reconnaissance in the area of the focused reconnaissance. The spatial region formed by the surrounding behavior may be used as a region of interest for the track. The surrounding shape in the region has ellipse, splay and the like, and analyzing the different shapes is helpful for identifying the behavior intention of the target object.
Currently, a common solution for typical array extraction of targets is to statistically analyze historical track data for a given target. Firstly, rasterizing track data, counting the number of track points and the number of track strips in different grids, screening out grids with higher heat, and clustering the grids according to the distance to obtain a matrix position area. According to the scheme, a part of array position areas with high heat and frequent movement can be extracted, however, for different flight targets, thresholds such as grid size, number of tracks in the grid, number of tracks and the like are required to be repeatedly adjusted, the problems of incomplete array position extraction, error array position extraction and the like exist, only discrete track points are reserved in the array position areas, time sequence characteristics are not existed, and the movement characteristics of the targets in the areas cannot be reflected.
Disclosure of Invention
Aiming at the problems existing in the prior art, the method for extracting the typical array position of the target is high in array position area recognition rate, good in calculation method robustness, capable of processing huge historical data, improving mining accuracy, reducing track analysis time and fast in robustness.
The technical scheme adopted by the invention is as follows: a method for extracting typical array positions of a massive track target comprises the following steps:
step 1, reading historical track point data from a historical track database;
step 2, mapping the read track point data into track data according to the track ID;
step 3, for any track, track smoothing, track compression and closed track extraction are sequentially carried out to obtain a compressed track and a closed track in the compressed track;
step 4, the tracks are summarized by taking the year and the object ID as the combined main key, and a compressed track set and a closed track set of each object in the corresponding year are obtained;
and 5, processing the compressed track set and the closed route set by adopting a clustering and array position combining mode aiming at the target of any year to obtain all array position areas and frequent active routes in the areas of the target.
In step 3, the method adopts the mode of outlier rejection and mean value smoothing to realize track smoothing.
Further, the specific method for eliminating the abnormal points comprises the following steps: and acquiring all track points in the track, sequentially calculating all track points except the head and tail track points and the adjacent track points to form an angle, and if the angle is smaller than a set threshold value, indicating that the track point is an abnormal track point, and deleting.
Further, the specific method for smoothing the average value is as follows: and designing a sliding window in the track, taking the average value of the positions of all track points in the window as the track point position in the middle of the window, sequentially and backwardly translating the sliding window until the whole track is traversed, and repeating the steps for three times to obtain a relatively smooth track curve.
Further, in the step 3, the specific steps of track compression are as follows:
step 3.1.1, initializing the track to be compressed DTR and the compressed track CTR into an original historical track TR and an empty set respectively
Sequentially judging whether track points in the DTR are reserved or not, and if so, adding the track points into the CTR;
step 3.1.2, calculating the distance from all track points to a line segment formed by the head point and the tail point in the track DTR to be compressed, and selecting the track point with the largest distance and the corresponding maximum distance;
step 3.1.3, if the maximum distance is smaller than or equal to the set track compression threshold, indicating that no track point in the track to be compressed DTR needs to be added into the compressed track CTR, and entering step 3.1.4; on the contrary, the track to be compressed is divided into two sections by taking the track point with the largest distance as a demarcation point, the first section of track is the track from the first track point to the track point with the largest distance, the second section of track is the track from the track point with the largest distance to the last track point, after the track point with the largest distance is added to the tail of the compressed track CTR, the first section of track and the second section of track are respectively used as new compressed tracks CTR from the step 3.1.2;
and 3.1.4, respectively adding the first track point and the last track point in the original historical tracks to the head and tail positions of the compressed track CTR to obtain the compressed track.
Further, the closed route extraction method comprises the following steps: and acquiring all track segments formed by adjacent track points in the compressed track, and if two non-adjacent track segments intersect, forming the track points between the two track segments into a closed track and storing the closed track.
Further, the specific steps of the step 5 include:
step 5.1, clustering track points and closed airlines in the compressed track set and the closed airlines set of any target respectively to obtain a track point cluster and a closed airline cluster;
and 5.2, selecting and combining to form a new array position area of the target according to the position relation between the track point cluster and the closed route cluster, and reserving the longest closed route as a frequent active route of the new array position area.
Further, in the step 5.1, the samples with similar distances are generalized into the same cluster by adopting a neighbor clustering mode; when the closed air route is clustered, the sample is the closed air route; the specific process of neighbor clustering is as follows:
step 5.1.1, setting a neighborhood radius epsilon, and calculating a sample set in a neighborhood of the sample p: epsilon Neighbors p ＝{p j ,dis(p,p j )<ε}，dis(p,p j ) Representing sample p j Distance from sample p;
step 5.1.2, detecting a sample p to be clustered which is not checked yet, if the sample p is not classified as a certain Cluster, establishing a new Cluster Cluster p Adding all samples in the neighborhood of p into a candidate set N;
step 5.1.3, searching samples q in the candidate set N, and classifying all samples in q and the neighborhood thereof into clusters Cluster p Adding all unprocessed samples in the neighborhood of q to a candidate set N;
step 5.1.4, repeating the step 5.1.3, and checking untreated samples in the candidate set N until all samples in the candidate set N are treated;
step 5.1.5, repeat step 5.1.2-step 5.1.4 until all objects fall into a cluster.
Further, when performing neighbor clustering on track points in a compressed track of a target, the Euclidean distance between samples from two track points; when the closed routes of the target are subjected to neighbor clustering, calculating the distance between samples by utilizing the ratio of the overlapping area between the closed areas formed by the two closed routes to the area.
Further, the substeps of the step 5.2 are as follows:
step 5.2.1, representing the track point clusters and the closed route clusters as circular areas;
step 5.2.2, taking a circular area set generated by the track points as a matrix candidate set, traversing the closed navigation line cluster, judging whether the closed navigation line cluster is intersected with a certain track point cluster, if so, merging the closed navigation line cluster with the track point cluster intersected with the closed navigation line cluster to generate a new matrix area, and replacing the merged candidate matrix in the matrix candidate set;
step 5.2.3, when combining the closed route cluster and the track point cluster, generating a new circular array position area by utilizing all track points in the cluster, and reserving the longest closed route in the closed route cluster as a frequent active route of a target in the array position area;
and 5.2.4, after all the closed route clusters and the track point clusters which need to be combined are combined, obtaining all the array position areas of the target and the frequent active routes in the corresponding areas.
Compared with the prior art, the beneficial effects of adopting the technical scheme are as follows:
1. the recognition rate of typical array bits is high, and the robustness of the calculation method is good. According to the method, the historical tracks of the target are read from the historical track library, a quick robust analysis model for array position extraction is established, different track shapes can be processed by means of closed-path extraction and the like, and the recognition accuracy of frequent track segments and typical array positions is improved; different smoothing algorithms are adopted, so that different abnormal track points can be processed, and the accuracy of array position extraction is ensured.
2. Typical array mining is fast. The method is realized based on the Spark framework, has the advantages of high calculation speed, high usability, wide universality, strong fusibility and the like, and has higher accuracy and timeliness; meanwhile, the historical tracks are compressed by adopting the Douglas-Peucker algorithm, so that the number of track points participating in subsequent calculation is reduced after compression, the calculated amount and the calculated time of subsequent steps are reduced, and the processing speed is further improved.
3. The invention can be used for the application of activity rule analysis, auxiliary identification and the like of moving targets such as airplanes, ships, automobiles and the like, can also be used for the cluster analysis of curves such as stocks, electrocardiograms and the like, and has stronger engineering practical value.
Drawings
Fig. 1 is a schematic diagram of a typical array position extraction method of a mass track target.
FIG. 2 is a flow chart of track analysis of a single track.
Fig. 3 is a flow chart of target bit extraction for a single target.
Detailed Description
Embodiments of the present application are described in detail below, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to like or similar modules or modules having like or similar functions throughout. The embodiments described below by referring to the drawings are exemplary only for the purpose of explaining the present application and are not to be construed as limiting the present application. On the contrary, the embodiments of the present application include all alternatives, modifications, and equivalents as may be included within the spirit and scope of the appended claims.
Aiming at the problems existing in the prior art, the embodiment provides a typical array position extraction method for massive track targets, which mainly comprises the steps of smoothly compressing tracks, and then merging track point clustering results by extracting frequently closed tracks to obtain array position areas, wherein the extracted array position areas comprise regions of interest and flight shapes in the regions, through which target objects frequently pass. The whole process in the embodiment is based on the Spark distributed data processing framework, the data processing is fully performed by utilizing the capacity of the clusters, the accuracy of data processing and the time efficiency of an algorithm are improved, the rapid analysis of massive track data is realized, the bottleneck of massive data generation is solved, and the extraction result can be analyzed and explored by professionals. The specific scheme is as follows:
as shown in fig. 1, a method for extracting typical array positions of a massive track target includes:
step 1, reading historical track point data from a historical track database;
step 2, mapping the read track point data into track data according to the track ID;
step 3, for any track, track smoothing, track compression and closed track extraction are sequentially carried out to obtain a compressed track and a closed track in the compressed track;
step 4, the tracks are summarized by taking the year and the object ID as the combined main key, and a compressed track set and a closed track set of each object in the corresponding year are obtained;
and 5, processing the compressed track set and the closed route set by adopting a clustering and array position combining mode aiming at the target of any year to obtain all array position areas and frequent active routes in the areas of the target.
The above process is described in detail herein:
for step 1, the history track is stored in the history track database in the form of a common track point, and in this embodiment { P } is adopted for the read history track point 1 ,P 2 ,…,P N And represents, where N is the number of historical track points, typically on the order of tens of millions,if the typical array position of the target is calculated directly based on the track point data, the calculation cost is quite high, so that a Map-Reduce algorithm is designed by utilizing a Spark platform in the subsequent processing process, and the subsequent step description can be seen specifically.
In step 2, the Map-Reduce algorithm is used to implement the Map process, i.e. the track data is mapped according to the track ID, see track point data. In the Map process, the historical track points { P } that are read are input 1 ,P 2 ,…,P N Output as track key value pair { (TRID) 1 ,TR 1 ),(TRID 2 ,TR 2 ),…(TRID n ,TR n )}，TRID n Representing track ID, TR n Representing the corresponding track; assume that the mapping results in n valid historical tracks { TR ] 1 ,TR 2 ,…,TR n More than or equal to 1, ith historical track TR i Is (x) j ,y j ),j＝1,2,…,n i ，n i In order to ensure the subsequent extraction effect, the i-th historical track is required to be the track point number n i ≥5,i＝1,2,…,n。
As shown in fig. 2, in step 3 in the present embodiment, each track in the track data sequentially performs track smoothing, track compression and closed route extraction, so as to obtain a corresponding compressed track and a closed route in the track; in particular, the method comprises the steps of,
since the real track is typically a smooth curve, the raw track data has many false error track points due to sensor errors or faults. Therefore, track smoothing processing needs to be performed on the original track data, and in this embodiment, track smoothing mainly includes two steps of outlier rejection and mean smoothing.
The abnormal point rejection is mainly realized based on angles formed by adjacent track points, because the flight speed of the aircraft is extremely fast, when turning and turning are needed, the aircraft cannot turn around immediately, the heading can only be gradually changed, and the generated track curve radian is changed less, and the method comprises the following specific steps:
step 3.1.1 for a single track TR, assuming the total number of track points is m, for all track points in the track except for the starting point and the ending pointP j J=2, …, -1, calculate P in the track j With track point P at the previous moment j-1 Formed track segment
And P j With track point P at the next moment j+1 Formed track section->
Is the inner product of:
step 3.1.2, calculating the track section
And track segment->
Is a modulus of the position coordinate vector of (c). The calculation formula is as follows.
Step 3.1.3, calculating the track point P j With fore-aft track point P j-1 、P j+1 Angle θ formed j 。
If theta is j Less than the angle threshold value ettheta=45°, the point P is judged j Is an abnormal track point, and is deleted.
In order to obtain a smooth track, the embodiment adopts a mean value smoothing algorithm to process after the abnormal point elimination is completed. The specific process is as follows: a sliding window (len=5 in this embodiment) is designed in the track, and the average value of all track point positions in the window is taken as the position of the track point in the middle of the window. The calculation formula is as follows:
and sequentially and backwardly translating the sliding window until the whole track is traversed, and repeating the steps for three times to obtain a relatively smooth track curve.
After the track smoothing is completed, the number of the obtained tracks is still huge, and the excessive number of track points of the tracks can cause excessive calculation amount of subsequent processing. Therefore, key points of the track are required to be extracted on the premise of keeping the shape of the track, the track is simplified, and the number of track points participating in subsequent calculation is reduced. The method comprises the following specific steps:
step 3.2.1, initializing the track to be compressed DTR and the compressed track CTR into an original historical track TR and an empty set respectively
Sequentially judging whether track points in the DTR are reserved or not, and if so, adding the track points into the CTR;
step 3.2.2, calculating all track points P in the track to be compressed DTR j Line segment P formed by two points from beginning to end 1 P m Distance dis (P) j ,P 1 P m ) J=1, 2, …, m, m is the track point number of DTR;
step 3.2.3, comparing the m track points P calculated in step 3.2.2 j To line segment P 1 P m M distances dis (P) j ,P 1 P m ) J=1, 2, …, m, selecting the point with the largest distance as j 1 The maximum distance is dis (P j1 ,P 1 P m )；
Step 3.2.4, judging the maximum distance dis (P j1 ,P 1 P m ) Whether the track compression threshold eCompressDis is greater than (ecompressdis=5 km in this embodiment), if so, it indicates that no track point in the DTR needs to be added to the CTR, otherwise, executing the subsequent steps;
step 3.2.5, 1 st to j th 1 The first section track of the point is used as a new track to be compressed DTR, and at the moment, the track point number of the DTR is m=j 1 Compressing the new track to be compressed DTR from the step 3.2.2;
step 3.2.6, j 1 Points are added from the tail to the compressed track CTR as retention points.
Step 3.2.7, j 1 The second section track from the point to the mth point is used as a new track to be compressed DTR, and at the moment, the track point number of the DTR is m=m-j 1 +1, compressing the new track to be compressed DTR from step 3.2.2.
And 3.2.8, respectively adding the 1 st point and the m th point to the head and tail positions of the compressed track CTR to obtain a final compressed track.
In this embodiment, the distance dis (P) in step 3.2.2 j ,P 1 P m ) The calculation method of (1) is as follows:
step 3.2.2.1 calculating Point P j To line segment P 1 P m Is of the drop foot P s The vertical foot coordinate position is expressed as (x s ,y s )：
If x 1 ＝x m X is then s ＝x 1 ,y s ＝y j 。
If y 1 ＝y m X is then s ＝x j ,y s ＝y 1 。
If x 1 ≠x m And y is 1 ≠y m By calculating line segment P 1 P m The slope k and intercept b give x s And y s ：
Step 3.2.2.2, calculating Point P j To line segment P 1 P m Distance dis (P) j ,P 1 P m )。
After the compressed track is obtained, the phenomenon that the aircraft targets spiral around the interest point to form a closed track is aimed at, and in the embodiment, the closed track in the track is further extracted through the intersection relation between track segments. The specific method comprises the following steps:
step 3.3.1 by
Representing the connection of adjacent track points P j And P j+1 The compressed track CTR may be represented as a track segment sequence ctr= { line j J=1, 2, …, -1}, m is the track point number of CTR;
step 3.3.2 for each track segment line in CTR j J=1, 2, …, -2, judging it and its non-adjacent track segment line k K=j+2, j+3, …, whether intersecting;
step 3.3.3 if two non-adjacent track segments line j And line k Intersecting to form a closed route curve of the track point between the two track sections jk ＝{P l Save =j+1, j+2, …, k-1} to closed-air-line set CCurves = { curve i ,＝1,2,3,…,K}。
In this embodiment, the line segment line in step 3.3.2 j And line k The specific judging method for whether the intersection is formed is as follows:
step 3.3.2.1, calculate line j And line k Forming an included angle
The calculation formula of (2) is as follows, and the detailed calculation steps refer to formulas (1) - (4).
Step 3.3.2.2, calculate line k Endpoint P of (2) k+1 To line j Distance dis (P) k+1 ,line j )＝dis(P k+1 ,P j P j+1 ) The detailed calculation steps refer to formulas (7) - (10).
Step 3.3.2.3, respectively setting an angle threshold and a distance threshold for judging whether the track segments intersect as follows: eThetaLine=90°, eDis=10 km, if
And dis (P) k+1 ,line j )<eDis, consider that two track segments intersect.
By adopting the proposed track smoothing, track compression and closed-path extraction methods to process all tracks to obtain corresponding compressed tracks and closed paths, a Map-Reduce algorithm is needed to realize a Reduce process, that is, the processing result of each track is summarized, in this embodiment, the year and the target ID are used as the combined main key, and the input of the Reduce process is the track analysis processing result set of a single track
{(TRID 1 ,(CTR 1 ,CCurves 1 )),(TRID 2 ,(CTR 2 ,CCurves 2 )),…(TRID n ,(CTR n ,CCurves n ) -n is the number of historical track stripes, TR) i And CCurves i Respectively, track TRID i The compressed tracks and the closed track sets in the tracks; the output is a key value { (Year) 1 ,MBID 1 ,(CTRs 11 ,CCurves 11 )),(Year 1 ,MBID 2 ,(CTRs 12 ,CCurves 12 )),…
(Year m ,MBID d ,(CTRs md ,CCurves md ) (v), m and p represent the total number of different year and goal types, CTRs, respectively jk And CCurves jk Respectively expressed in Year as Year j When the target ID is MBID k A compressed track set and a closed path set of targets of (1).
The interest area of the track after track compression may be split into a plurality of discrete areas, so that closed-route extraction and clustering are required to be performed on the compressed track to retain frequent closed-route information in original data, candidate array positions are obtained by clustering compressed track points, and the array positions are combined through the frequent closed-route, so that more accurate array position areas and active routes in the areas can be obtained. As shown in fig. 3, the specific procedure is as follows:
step 5.1, clustering the track points and the closed airlines respectively to obtain frequent active areas and frequent closed airlines;
and 5.2, merging part of the array position areas into a larger array position area by using the frequent closed route, namely, the target array position area, and taking the closed route with the largest track point number as the frequent active route.
In this embodiment, in the step 5.1, the clustering of the track points and the closed route is adjacent clustering, that is, the samples with similar distances are summarized into the same cluster, and the specific method is as follows:
step 5.1.1, designated distance threshold esampleidis=20 km. Setting the neighborhood radius epsilon=esampleidis, then the epsilon-neighborhood of the sample p is defined as the set of samples with radius epsilon centered on p.
εNeighbors p ＝{p j ,dis(p,p j )<ε} (12)
Wherein dis (p, p j ) Representing the inter-sample distance;
step 5.1.2, detecting an object p to be clustered which is not checked yet, if the p is not classified as a certain Cluster, establishing a new Cluster Cluster p Adding all samples in the epsilon-neighborhood of pA candidate set N;
step 5.1.3, searching the object q of the candidate set N, and classifying all q and neighborhood objects thereof into clusters Cluster p All unprocessed samples in the epsilon-neighborhood of q are added to candidate set N.
Step 5.1.4, repeating step 5.1.3, and continuing to check the unprocessed objects in N until the current candidate set N is empty.
Step 5.1.5, repeat step 5.1.2-step 5.1.4 until all objects fall into a cluster.
And 5.1.6, counting the number of samples in each cluster, calculating to obtain the average number of samples as a threshold eConnt, and screening clusters with the number of samples in the cluster larger than the threshold as clustering results.
Based on the proximity clustering method, for track points in all compressed tracks of the target, euclidean distance between two track points is used as the distance between samples, namely
In order to simplify the calculation and increase the processing speed for the closed route of the target, in this embodiment, the distance between samples is calculated by the ratio of the overlapping area between the closed areas formed by the two closed routes to the area, and the higher the area ratio, the smaller the distance between the two closed route samples is, namely:
the difficulty of directly calculating the area of a closed region formed by a closed route is high, so that the closed region is approximately expressed as a rectangular region curve j Matrix { xmin, ymin, xmax, ymax }, xmin, xmax, ymin, ymax are the closed course curve, respectively j The minimum longitude, the maximum longitude, the minimum latitude and the maximum latitude of all track points in the closed area are calculated according to the following formula.
Two closed route curve j And cut ve s Is of the intersection area of (2)
The calculation formula is as follows.
x＝(min(xmax j ,xmax s )-max(xmin j ,xmin s ))
y＝(min(ymax j ,ymax s )-max(ymin j ,ymin s )) (16)
In the step 5.2, the merging method of the array position areas specifically comprises the following steps:
and 5.2.1, representing the track point clusters and the closed route clusters obtained by clustering in the step 5.1 as circular areas. Traversing all track points in the track point cluster or the closed track cluster to obtain maximum and minimum values of longitude and latitude, wherein the maximum and minimum values are respectively expressed as xmax, xmin, ymax and ymin, and the circular area can be expressed as:
Circle＝{O,r}
and 5.2.2, taking a circular area set generated by the track point cluster as a matrix candidate set. Traversing the closed navigation line cluster, if the closed navigation line cluster is intersected with a certain track point cluster, combining the closed navigation line cluster with the track point cluster intersected with the closed navigation line cluster to generate a new array position area, and replacing the combined candidate array position in the array position candidate set. Whether the closed course cluster and the course point cluster intersect the distance between the circular areas formed by the closed course cluster and the course point cluster is judged. The calculation formula is as follows.
If the distance dis (Cluster) j ,Cluster s ) Less than the matrix distance threshold ekkeyarea = 20km, then consider intersection.
And 5.2.3, for the closed route clusters and the track point clusters which need to be combined, generating a new circular array position area according to the step 5.2.1 by all track points in the clusters, and reserving the longest closed route in the closed route clusters, namely the closed route with the largest track point number, as an intra-area representative curve for describing the frequent moving route of the target in the array position area.
At this time, a more accurate target array position area and an in-area active route are obtained, and the method is realized based on a Spark distributed data processing framework, so that the data processing performance of a large data platform can be effectively improved, the problem of insufficient performance during data analysis is solved by utilizing an RDD calling mechanism and Spark memory computing capacity, and the rapid processing analysis requirement of mass data can be effectively solved.
It should be noted that, in the description of the embodiments of the present invention, unless explicitly specified and limited otherwise, the terms "disposed," "connected," and "connected" should be interpreted broadly, and may be, for example, fixedly connected, detachably connected, or integrally connected; may be directly connected or indirectly connected through an intermediate medium. The specific meaning of the above terms in the present invention will be understood in detail by those skilled in the art; the accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Although embodiments of the present application have been shown and described above, it will be understood that the above embodiments are illustrative and not to be construed as limiting the application, and that variations, modifications, alternatives, and variations may be made to the above embodiments by one of ordinary skill in the art within the scope of the application.
Claims (10)
1. A method for extracting typical array positions of a massive track target is characterized by comprising the following steps:
step 1, reading historical track point data from a historical track database;
step 2, mapping the read track point data into track data according to the track ID;
step 3, for any track, track smoothing, track compression and closed track extraction are sequentially carried out to obtain a compressed track and a closed track in the compressed track;
step 4, the tracks are summarized by taking the year and the object ID as the combined main key, and a compressed track set and a closed track set of each object in the corresponding year are obtained;
and 5, processing the compressed track set and the closed route set by adopting a clustering and array position combining mode aiming at the target of any year to obtain all array position areas and frequent active routes in the areas of the target.
2. The method for extracting typical array positions of a massive track target according to claim 1, wherein in the step 3, the track smoothing is realized by adopting a mode of outlier rejection and mean smoothing.
3. The method for extracting typical array positions of a mass track target according to claim 2, wherein the specific method for eliminating abnormal points is as follows: and acquiring all track points in the track, sequentially calculating all track points except the head and tail track points and the adjacent track points to form an angle, and if the angle is smaller than a set threshold value, indicating that the track point is an abnormal track point, and deleting.
4. The method for extracting typical array positions of a mass track target according to claim 2, wherein the specific method for mean smoothing is as follows: and designing a sliding window in the track, taking the average value of the positions of all track points in the window as the track point position in the middle of the window, sequentially and backwardly translating the sliding window until the whole track is traversed, and repeating the steps for three times to obtain a relatively smooth track curve.
5. The method for extracting typical array positions of a massive track target according to claim 1, wherein in the step 3, the specific steps of track compression are as follows:
step 3.1.1, initializing the track to be compressed DTR and the compressed track CTR into an original historical track TR and an empty set respectively
Sequentially judging whether track points in the DTR are reserved or not, and if so, adding the track points into the CTR;
step 3.1.2, calculating the distance from all track points to a line segment formed by the head point and the tail point in the track DTR to be compressed, and selecting the track point with the largest distance and the corresponding maximum distance;
step 3.1.3, if the maximum distance is smaller than or equal to the set track compression threshold, indicating that no track point in the track to be compressed DTR needs to be added into the compressed track CTR, and entering step 3.1.4; on the contrary, the track to be compressed is divided into two sections by taking the track point with the largest distance as a demarcation point, the first section of track is the track from the first track point to the track point with the largest distance, the second section of track is the track from the track point with the largest distance to the last track point, after the track point with the largest distance is added to the tail of the compressed track CTR, the first section of track and the second section of track are respectively used as new compressed tracks CTR from the step 3.1.2;
and 3.1.4, respectively adding the first track point and the last track point in the original historical tracks to the head and tail positions of the compressed track CTR to obtain the compressed track.
6. The method for extracting typical array positions of a mass track target according to claim 1, wherein the method for extracting the closed route is as follows: and acquiring all track segments formed by adjacent track points in the compressed track, and if two non-adjacent track segments intersect, forming the track points between the two track segments into a closed track and storing the closed track.
7. The method for extracting typical array positions of a mass track target according to claim 1, wherein the specific steps of the step 5 include:
step 5.1, clustering track points and closed airlines in the compressed track set and the closed airlines set of any target respectively to obtain a track point cluster and a closed airline cluster;
and 5.2, selecting and combining to form a new array position area of the target according to the position relation between the track point cluster and the closed route cluster, and reserving the longest closed route as a frequent active route of the new array position area.
8. The method for extracting typical array positions of mass track targets according to claim 7, wherein in the step 5.1, samples with similar distances are induced into the same cluster by adopting a neighbor clustering mode; when the closed air route is clustered, the sample is the closed air route; the specific process of neighbor clustering is as follows:
step 5.1.1, setting a neighborhood radius epsilon, and calculating a sample set in a neighborhood of the sample p: epsilon Neighbors p ＝{p j ，dis(p，p j )<ε}，dis(p，p j ) Representing sample p j Distance from sample p;
step 5.1.2, detecting a sample p to be clustered which is not checked yet, if the sample p is not classified as a certain Cluster, establishing a new Cluster Cluster p Adding all samples in the neighborhood of p into a candidate set N;
step 5.1.3, searching samples q in the candidate set N, and classifying all samples in q and the neighborhood thereof into clusters Cluster p Adding all unprocessed samples in the neighborhood of q to a candidate set N;
step 5.1.4, repeating the step 5.1.3, and checking untreated samples in the candidate set N until all samples in the candidate set N are treated;
step 5.1.5, repeat step 5.1.2-step 5.1.4 until all objects fall into a cluster.
9. The method for extracting typical array positions of a massive track target according to claim 8, wherein when performing neighbor clustering on track points in a compressed track of the target, the inter-sample distances are the euclidean distances of two track points; when the closed routes of the target are subjected to neighbor clustering, calculating the distance between samples by utilizing the ratio of the overlapping area between the closed areas formed by the two closed routes to the area.
10. The method for extracting typical array positions of a mass track target according to claim 7, wherein the sub-steps of the step 5.2 are as follows:
step 5.2.1, representing the track point clusters and the closed route clusters as circular areas;
step 5.2.2, taking a circular area set generated by the track points as a matrix candidate set, traversing the closed navigation line cluster, judging whether the closed navigation line cluster is intersected with a certain track point cluster, if so, merging the closed navigation line cluster with the track point cluster intersected with the closed navigation line cluster to generate a new matrix area, and replacing the merged candidate matrix in the matrix candidate set;
step 5.2.3, when combining the closed route cluster and the track point cluster, generating a new circular array position area by utilizing all track points in the cluster, and reserving the longest closed route in the closed route cluster as a frequent active route of a target in the array position area;
and 5.34, after all the closed route clusters and the track point clusters which need to be combined are combined, obtaining all the array position areas of the target and the frequent active routes in the corresponding areas.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211623673.3A CN116244391A (en) | 2022-12-16 | 2022-12-16 | Method for extracting typical array position of massive track targets |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211623673.3A CN116244391A (en) | 2022-12-16 | 2022-12-16 | Method for extracting typical array position of massive track targets |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116244391A true CN116244391A (en) | 2023-06-09 |
Family
ID=86633907
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211623673.3A Pending CN116244391A (en) | 2022-12-16 | 2022-12-16 | Method for extracting typical array position of massive track targets |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116244391A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117251748A (en) * | 2023-10-10 | 2023-12-19 | 中国船舶集团有限公司第七〇九研究所 | Track prediction method, equipment and storage medium based on historical rule mining |
-
2022
- 2022-12-16 CN CN202211623673.3A patent/CN116244391A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117251748A (en) * | 2023-10-10 | 2023-12-19 | 中国船舶集团有限公司第七〇九研究所 | Track prediction method, equipment and storage medium based on historical rule mining |
| CN117251748B (en) * | 2023-10-10 | 2024-04-19 | 中国船舶集团有限公司第七〇九研究所 | Track prediction method, equipment and storage medium based on historical rule mining |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Himstedt et al. | Large scale place recognition in 2D LIDAR scans using geometrical landmark relations | |
| CN110188093A (en) | A kind of data digging system being directed to AIS information source based on big data platform | |
| CN110609881A (en) | Vehicle trajectory deviation detection method, system and storage medium | |
| CN107679558A (en) | A kind of user trajectory method for measuring similarity based on metric learning | |
| KR20140043393A (en) | Location-aided recognition | |
| CN107832778B (en) | Same target identification method based on spatial comprehensive similarity | |
| CN112633389B (en) | Hurricane movement track trend calculation method based on MDL and speed direction | |
| Jung et al. | Noisy and incomplete fingerprint classification using local ridge distribution models | |
| CN115761534A (en) | Method for detecting and tracking small target of infrared unmanned aerial vehicle under air background | |
| CN116244391A (en) | Method for extracting typical array position of massive track targets | |
| Yin et al. | Pse-match: A viewpoint-free place recognition method with parallel semantic embedding | |
| Li et al. | Exploring label probability sequence to robustly learn deep convolutional neural networks for road extraction with noisy datasets | |
| Men et al. | Airport runway semantic segmentation based on DCNN in high spatial resolution remote sensing images | |
| CN117251748B (en) | Track prediction method, equipment and storage medium based on historical rule mining | |
| Qiao et al. | Rapid trajectory clustering based on neighbor spatial analysis | |
| CN111221819B (en) | Similar typhoon rapid matching method based on multidimensional data space segmentation | |
| Jiang et al. | A fast-mining method for target behavior pattern based on trajectory data | |
| CN107389071B (en) | Improved indoor KNN positioning method | |
| CN116484244A (en) | Automatic driving accident occurrence mechanism analysis method based on clustering model | |
| CN115862385A (en) | Airport terminal area aircraft flight mode mining method based on standard flight program | |
| Wang et al. | Grid‐Based Whole Trajectory Clustering in Road Networks Environment | |
| Zhuang et al. | Aerial battlefield targets grouping based on DTW-DBSCAN algorithm | |
| Yin et al. | Multi-scale graph convolutional network for intersection detection from gps trajectories | |
| Gao et al. | A Novel Semantic Representation of Airport Surface Trajectory for Taxiing Pattern Recognition | |
| CN110727749A (en) | Self-adaptive candidate road section searching method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |