CN109831212A

CN109831212A - A kind of time locus expression and compression frame that can make full use of data characteristics

Info

Publication number: CN109831212A
Application number: CN201910146589.9A
Authority: CN
Inventors: 陈超; 赵杰; 丁琰
Original assignee: Chongqing University
Current assignee: Chongqing University
Priority date: 2019-02-27
Filing date: 2019-02-27
Publication date: 2019-05-31
Anticipated expiration: 2039-02-27
Also published as: CN109831212B

Abstract

The present invention provides a kind of time locus expressions and compression frame that can make full use of data characteristics, are related to the compression of trace compression field, especially time locus.A large amount of and redundancy track of vehicle data result in many data storages, the expense of communication and processing aspect.Currently, initial trace is often decomposed into the time series of space path sum.Compression herein for time series proposes a set of new frame, including an expression constructor and four compressors.Specifically, we have constructed a tool first, there are four types of the time locus of element to indicate, it has lower entropy and storage；Then four kinds of compressors are constructed and compress corresponding four dvielement respectively, and improve compression efficiency.

Description

A kind of time locus expression and compression frame that can make full use of data characteristics

Technical field

The present invention relates to trace compression technical fields, especially relate to time locus compress technique.

Background technique

A large amount of track datas that mobile object generates will lead to data storage, the expense of communication and processing.Track of vehicle number According to being one of typical case.But since vehicle can only travel on road network, limited with special space.In order to subtract The light above problem, reducing its size while effectiveness (as supported inquiry) for keeping track data is a kind of very promising side Method.Track is one group of GPS point sequence, records the space time information of move vehicle, by timestamp t_i, geospatial coordinates (x_i,y_i) and Instantaneous velocity v_iComposition.Currently, being often decomposed into the road indicated by a series of continuous side by the track that move vehicle generates A series of space path in road network and associated by timestamp t_iThe time series of expression.Decomposing trajectories ensure There is better compression performance in terms of validity, this is proven by theory analysis.In addition, more preferable in addition to obtaining Compression performance outside, the separation of room and time is also beneficial to inquire.For example, a common inquiry is that vehicle is when specific Between traveling on which road.

In terms of space tracking compression, it has been proposed that many outstanding algorithms.However in terms of time locus compression, It works with having and does not well solve problem.Firstly, good time locus expression should have small entropy (entropy) and size (size).The smaller explanation expression of entropy contains a large amount of duplicate messages, helps further to encode compression.In addition, concision and compact Expression facilitate reduce carrying cost.Unfortunately, for the two are required, existing work is underproof.Secondly, Current expression cannot support some inquiries, because they have ignored some temporal informations, especially acceleration and instantaneous velocity. These information are most important for many location based services (LBS).For example, acceleration information is in research driving behavior classification In indispensable one of factor.Instantaneous velocity can capture fine-grained driving situation, this is for the behavioral value system that exceeds the speed limit It is extremely important.Finally, existing many representational compression methods, such as PRESS, CCF and TED etc. all can only be biggish Good compression effectiveness is realized in error range.

Summary of the invention

To solve the above problems, indicating and compressing the present invention provides a kind of time locus that can make full use of data characteristics Frame constructs a kind of new time locus expression, and is effectively compressed.

Specifically, the present invention provides a kind of time locus that can make full use of data characteristics to indicate and compression frame Scheme are as follows:

A kind of time locus expression and compression frame that can make full use of data characteristics, includes two stages.Wherein, first A stage constructs a kind of new time locus expression, including 4 kinds of components: initial information, time series, auxiliary information sequence Column and acceleration degree series；Second stage, for each above-mentioned component, we indicate that feature extraction is a kind of using it Basic compressor, and compression efficiency is further increased using its potential low-level image feature.

Further, a kind of time locus that can make full use of data characteristics of the present invention indicates and first in compression frame A stage includes following 4 steps: step 1: in j-th strip initial trace section Tr_jIn, by initial time st_jAnd starting velocity sv_jIt is saved as initial information, i.e. II_j=(st_j；sv_j)；Step 2: in j-th strip initial trace section Tr_jIn, it constructs corresponding Time series TS_j=< Δ t₁；Δt₂；…；Δt_l-1>, wherein Δ t_iIt is time interval；Step 3: in j-th strip initial trace section Tr_jIn, acceleration sequence is expressed asWherein1≤i≤l-1, the sequential recording Average acceleration and instantaneous velocity in section (instantaneous velocity can be combined by initial information and average acceleration to be acquired)；Step 4: in j-th strip initial trace section Tr_jIn, auxiliary information sequence is expressed as AUXS_j=< aux₁；…；aux_l-1>, wherein 1≤i≤l-1。

Further, a kind of time locus that can make full use of data characteristics of the present invention indicates and second in compression frame A stage includes following 4 steps: step 1: building initial information compressor, with initial time st_jjFor sequence, the is selected One moment is as fiducial time t_b, then only need to record time interval stb_j(i.e.st_j-t_b), to reduce time letter The digit of breath, reduces storage overhead；Step 2: building Time Series Compression device, for time series TS_j=< Δ t₁；Δ t₂；…；Δt_l-1>, it is concentrated in quality data, almost all of time interval Δ t_iIt is equal to the sampling time Δ T of GPS point, It is compressed at this time using run length encoding；Δ t is concentrated in low quality data_iIn floating up and down for Δ T value, at this time using Kazakhstan Fu Man coding is compressed；Step 3: building acceleration sequence compressor obtains first by the value discretization of accelerationHuffman encoding is reused to be compressed；Step 4: building auxiliary information sequence compressor, First by the value discretization of auxiliary information, AUXS is obtained_j=< D (aux₁)；…；D(aux_l-1) >, reuses Huffman encoding progress Compression.

Detailed description of the invention

Fig. 1 is system framework figure of the invention；

Fig. 2 is the new expression that sample initial trace segment constructs；

Fig. 3 is the schematic diagram of Huffman encoding；

Fig. 4 is the frequency distribution of the acceleration value of different zones.

Specific embodiment

The following further describes the present invention with reference to the drawings.

Before introducing the content of present invention, 4 necessary concepts in the present invention are first introduced.

1st concept: road network, road network are a figure G (N, E), include an one group of line set E and group node set N.In E Each element is a directed edge e_i, it is associated with two nodes.Each node n is in N by the seat of a pair of of longitude and latitude Mark combination indicates, corresponds to the spatial position of sampled point.

2nd concept: initial trace segment, initial trace segment Tr_jIt is a series of GPS track points, is denoted as Tr_j=< p_i, p_i+1,…,p_i+l-1>, one of GPS track point p_iThe space time information for having recorded vehicle, is denoted as p_i=(t_i,lat_i,lon_i,v_i), Parameter l controls the length of segment.

3rd concept: the average acceleration between two continuous GPS pointsIt is denoted as1≤i≤l-1。

4th concept: auxiliary information aux_i, for saving the road network range information between two continuous GPS points. 1≤i≤l-1, wherein Dist (x, y) is the function for calculating road network distance.

A kind of time locus expression and compression frame that can make full use of data characteristics, takes conjunction to initial trace information The representation of reason, and then carry out effective compression processing.Frame includes two stages, as shown in Figure 1, first stage is Some initial trace fragment datas are given, indicate that constructor constructs a kind of new expression by time locus；Second stage To be indicated for the time locus constructed, different components is compressed using different compressors respectively.

1, first stage includes following four step:

Step 1: in j-th strip initial trace segment Tr_jIn, by initial time st_jJ and starting velocity sv_jAs initial information It saves, i.e. II_j=(st_jj；sv_j)；

Step 2: in j-th strip initial trace section Tr_jIn, construct corresponding time series TS_j=< Δ t₁；Δt₂；…；Δ t_l-1>, wherein

Step 3: in j-th strip initial trace section Tr_jIn, acceleration sequence is expressed asIts In1≤i≤l-1, (instantaneous velocity can be by for average acceleration and instantaneous velocity in sequential recording section Initial information and average acceleration, which are combined, to be acquired)；

Step 4: in j-th strip initial trace section Tr_jIn, auxiliary information sequence is expressed as AUXS_j=< aux₁；…； aux_l-1>, wherein1≤i≤l-1。

By the construction of first stage, the new time locus of available one kind is indicated.As shown in Fig. 2, with an original Beginning path segment Tr₁=< p₁,p₂,p₃,p₄> for, being inputted time locus indicates constructor, obtains new being expressed as II₁= (st₁；sv₁), TS₁=< Δ t₁, Δ t₂, Δ t₃>,AUXS₁=< aux₁, aux₂, aux₃>。

2, second stage includes following four step:

Step 1: building initial information compressor, such as initial time sequence st_1,2,3,4,5=(1545188400, 1545192000,1545195600,1545199200,1545202800) first moment 1545188400, is selected as benchmark Time t_b, then only record time interval stb_j(i.e.st_j-t_b), then st_1,2,3,4,5Be converted to stb_1,2,3,4,5=(3600, 7200,10800,14400,18000), and then reduce the digit of temporal information, finally, the considerations of for information decoding, we It is additionally arranged length mark position character string, identifies and stores each stb_jEnd position；

Step 2: building Time Series Compression device, for time series TS_j=< Δ t₁；Δt₂；…；Δt_l-1>, consider with Lower two kinds of situations.It is concentrated in quality data, almost all of time interval Δ t_iAll it is identical, and is equal to adopting for GPS point Sample time Δ T is compressed using run length encoding at this time, such as TS_j=<6,6,6,6,6,6,6,6,6,7>, after compression Result be1001011000010111 (9 6 and 17, binary representation)；It is concentrated in low quality data, because of GPS signal The relationship of delay makes Δ t_iIn floating up and down for Δ T value, compressed at this time using Huffman encoding, specific example is with reference to figure 3；

Step 3: building acceleration sequence compressor, forIt will be accelerated using formula (1) first DegreeDiscretization obtains

Then Huffman encoding pair is usedIt is compressed.To further improve compression efficiency, it is contemplated that different zones Acceleration frequency distribution has different, as shown in Figure 4.So more Huffman trees can be constructed for different zones, together When occupy memory space excessively in order to avoid Huffman tree number, finally used Huffman tree folding；

Step 4: building auxiliary information sequence compressor, for AUXS_j=< aux₁；…；aux_l-1>, formula is used first (2) by auxiliary information aux_iDiscretization obtains AUXS_j=< D (aux₁)；…；D(aux_l-1)>

Then Huffman encoding pair is usedIt is compressed, finally further increases compression efficiency, due to D (aux_i) have A large amount of not repetition values, cause the Huffman tree number of plies constructed more, so that cataloged procedure is quite time-consuming, it is contemplated that traveling speed Degree is to D (aux_i) influence significantly, therefore lesser Huffman tree can be respectively constructed on the road of different speed limit types in terms of reducing Calculate expense.

Claims

1. a kind of time locus that can make full use of data characteristics indicates and compression frame, it is characterised in that include following two rank Section: (1) in the first stage, constructing a kind of new time locus indicates, including 4 kinds of components: initial information, time series, Auxiliary information sequence and acceleration degree series；

(2) in second stage, for each above-mentioned component, we indicate a kind of basic compression of feature extraction using it Device, and compression efficiency is further increased using its potential low-level image feature.

2. a kind of time locus that can make full use of data characteristics according to claim 1 indicates and compression frame, special Sign is that the first stage includes following four step:

Step 1: in j-th strip initial trace segment Tr_jIn, by initial time st_jWith starting velocity sv_jIt is saved as initial information, That is II_j=(st_j；sv_j)；

Step 2: in j-th strip initial trace section Tr_jIn, construct corresponding time series TS_j=< Δ t₁；Δt₂；…；Δt_l-1>, Wherein

Step 3: in j-th strip initial trace section Tr_jIn, acceleration sequence is expressed asWherein (instantaneous velocity can be by first for average acceleration and instantaneous velocity in sequential recording section Beginning information and average acceleration, which are combined, to be acquired)；

Step 4: in j-th strip initial trace section Tr_jIn, auxiliary information sequence is expressed as AUXS_j=< aux₁；…；aux_l-1>, Wherein

3. a kind of time locus that can make full use of data characteristics according to claim 1 indicates and compression frame, special Sign is that the second stage includes following four step:

Step 1: building initial information compressor, such as initial time sequence, first moment is selected as fiducial time t_b, then only record time interval stb_j(i.e.st_j-t_b), to reduce the digit of temporal information, in addition, for information solution The considerations of code, we are additionally arranged length mark position character string；

Step 2: building Time Series Compression device, for time series TS_j=< Δ t₁；Δt₂；…；Δt_l-1>, consider following two Kind situation.It is concentrated in quality data, almost all of time interval Δ t_iAll be identical, and be equal to GPS point sampling when Between Δ T, compressed at this time using run length encoding；It is concentrated in low quality data, because the relationship of GPS signal delay makes Δt_iIn floating up and down for Δ T value, compressed at this time using Huffman encoding；

Step 3: building acceleration sequence compressor, forUse formula (1) by acceleration first Discretization obtains

Wherein δ is level of discretization --- (1)

Then Huffman encoding pair is usedIt is compressed, it is contemplated that it is certain poor that the acceleration frequency distribution of different zones has Different, we construct more Huffman trees for different zones to improve compression efficiency, while in order to avoid Huffman tree number mistake It occupies memory space more, has finally used Huffman tree folding；

Step 4: building auxiliary information sequence compressor, for AUXS_j=< aux₁；…；aux_l-1>, use formula (2) will first Auxiliary information aux_iDiscretization obtains AUXS_j=< D (aux₁)；…；D(aux_l-1)>

Then Huffman encoding pair is usedIt is compressed, finally further increases compression efficiency, due to D (aux_i) have greatly Unique value is measured, the Huffman tree number of plies constructed is more, coding time, it is contemplated that travel speed is to D (aux_i) influence significantly, therefore Lesser Huffman tree can respectively be constructed on the road of different speed limit types to reduce computing cost.