CN106355881A - Space-autocorrelation-based traffic accident blackspot identification method and device - Google Patents

Abstract

The invention belongs to the technical field of road safety evaluation, and discloses a space-autocorrelation-based traffic accident blackspot identification method and a space-autocorrelation-based traffic accident blackspot identification device. A geographical distribution characteristic is described from the angle of geographical accident distribution by combining accident attribute values on road space units on the premise of a proximity relationship between the road space units, and a traffic accident blackspot is identified by adopting a kernel density estimation and space autocorrelation method. The shortcoming of relatively poorer visual expression capability of research data due to adoption of a classical mathematical statistical analysis method for a traffic accident blackspot identification process in the prior art is overcome. By the method, traffic accident blackspots at road sections and intersections can be identified, so that identification results are more accurate. In addition, a special analysis function is fully utilized on the basis of spatial position and geographical position information, and a foundation is laid for finding out accident causes.

Description

A kind of hazardous location identification method and device based on spatial autocorrelation

Technical field

The invention belongs to road safety assessment technical field, it is related to a kind of hazardous location identification based on spatial autocorrelation Method and device.

Background technology

The discriminating of Accident Area is road safety assessment and the weight targetedly traffic safety measure being improved Want premise, conventional hazardous location identification method have accident number method, accident rate method, equivalent total accident number method, cluster analyses and Empirical Bayes method etc..

Accident number method and accident rate method are intended to determine that a marginal value, as the index judging Accident Area, is applied to public affairs Road, does not consider the impact of random fluctuation, is not suitable for differentiating on a large scale；Matrix method overcomes accident number method to a certain extent With the defect of accident rate method, comprehensive accident number method and accident rate method are ranked up to tested point and differentiate, but its criterion There is certain subjectivity, still can not process the problem of random fluctuation；Gray scale assessment method is based on gray theory, and criterion is still There is certain subjectivity；Quality control method, using accident rate as identification beacon, removes evaluation index according to Principle of Statistics Contiguous space, needs a large amount of historical datas in similar place, China's data information at this stage does not still satisfy the requirements, in addition, work Work amount is big；Equivalent accident sum method be based on microcosmic traffic analysis non-statistical safety evaluation method although not relying on accident Statistical data, but investigation amount is huge, portable poor；Classical Bayesian Method considers the randomness of accident, improves prediction Accuracy, but data acquisition amount is big, procedure operation is complicated.

During above-mentioned hazardous location identification analysis, substantially rely on classical Mathematical Statistic Analysis Method and model, research The result form of expression is single, and data visualization ability to express is poor；Typically crossing is evaluated respectively with section, have ignored both Influence each other；Additionally, ignoring locus and the geographical attribute information of accident generation, spatial analysis functions are not enough, to potential thing Therefore the excavating depth of reason is inadequate.

Content of the invention

It is an object of the invention to provide a kind of road accident Multiple trauma discrimination method based on spatial autocorrelation, overcome existing Differentiating in technology is had during Accident Area, to differentiate factor defect not comprehensively, inadequate to the excavation of latent defect reason depth.

For reaching above-mentioned purpose, the technical solution used in the present invention is:

A kind of road accident Multiple trauma discrimination method based on spatial autocorrelation, comprises the steps:

Two dimensional Distribution using the accident point to be detected in nonparametric probability model representation region to be measured；

Determine the optimal bandwidth h of described Density Estimator model₀；

Spatial statistical units are built using " face-arc " network topology model based on geosystem；

Judge whether the incident attributes on described spatial statistical units assume spatial autocorrelation；

If the incident attributes on described spatial statistical units assume spatial autocorrelation, calculate described spatial statistical units The distribution results of standardization z (i) value of local statistic；

The distribution results of standardization z (i) value according to described local statistic generate the scattergram of Accident Area.

The above-mentioned road accident Multiple trauma discrimination method based on spatial autocorrelation, wherein, described Density Estimator model Kernel density function is:

$\hat{u}\left(x\right)=\frac{1}{{\mathrm{nh}}^d}{\mathrm{\σ}}_{i=1}^{n}k\left(\frac{l-l_i}{h}\right);$

Wherein, k () is kernel function, and h is window width, and n is the number of the accident point to be measured in region to be measured, and d is the dimension of data Number, (l-l_i) for accident point to be measured to accident point l_iDistance.

The above-mentioned road accident Multiple trauma discrimination method based on spatial autocorrelation, wherein, the described cuclear density letter of described determination The optimal bandwidth of number, comprising:

Obtain real density f (x) in described region to be measured for the described accident point to be measured；

Obtain described kernel density function and the corresponding window width during integrated square error minimum of described real density f (x) h₀, as optimal bandwidth.

The above-mentioned road accident Multiple trauma discrimination method based on spatial autocorrelation, wherein, described kernel density function with described The integrated square error computing formula of real density f (x) is:

WhereinFor described kernel density function,For institute State integrated square error.

The above-mentioned road accident Multiple trauma discrimination method based on spatial autocorrelation, wherein, described employing is based on geosystem " face-arc " network topology model build spatial statistical units, comprising:

Region to be measured is divided into intersection region and section region；

Accident point in the region of intersection, sets up the sky of intersection and accident by the operational rule in " face " and " point " Between relation；

The accident point in section region, overlaps, by spatial data, the spatial relationship that road and accident are set up in computing, and Described by spatial data overlap budget set up before spatial relationship, in GIS-Geographic Information System (geographic information System, abbreviation gis) in set the relief area of predetermined width for segmental arc.

The above-mentioned road accident Multiple trauma discrimination method based on spatial autocorrelation, wherein, the described spatial statisticses of described judgement Whether the incident attributes on unit assume spatial autocorrelation, comprising:

Step s41, calculates autocorrelation exponent i, expectation e (i) and expectation e (i) of incident attributes and the standard of variance construction Change z₁Value；

Step s42, judges whether the autocorrelation exponent i of described incident attributes is more than zero, and if so, statistic unit is then described The incident attributes of adjacent domain are in space positive correlation, continue step s43, otherwise, exit this operation；

Step s43, the standardization z that the expectation of described autocorrelation exponent is constructed with variance₁Value be more than preset value when, represent On statistic unit, the incident attributes in the region of described Spatial Adjacency or spatial neighbor are in spatial autocorrelation.

The above-mentioned road accident Multiple trauma discrimination method based on spatial autocorrelation, wherein, the calculating of described autocorrelation exponent i Method is:

$i=\frac{{\mathrm{n\σ\σw}}_{ij\left({\mathrm{sp}}_{i-}\stackrel{\&overbar;}{s}p\right)\left({\mathrm{sp}}_{j-}\stackrel{\&overbar;}{s}p\right)}}{\left({\mathrm{\σ\σw}}_{ij}\right)\left(\mathrm{\σ}{\left({\mathrm{sp}}_{i-}\stackrel{-}{sp}\right)}^2\right)};$

Wherein, w_ijFor the Spatial weight matrix of position relationship between space cell i and j, sp_i, sp_jFor space cell i and j The accident number at place or severity of injuries；For the meansigma methodss of accident number or severity of injuries, n is space cell sum.

The above-mentioned road accident Multiple trauma discrimination method based on spatial autocorrelation, wherein, judges described space system described Before whether the incident attributes on meter unit assume spatial autocorrelation, also include:

Determine that incident attributes are severity of injuries and/or accident number.

The above-mentioned road accident Multiple trauma discrimination method based on spatial autocorrelation, wherein, the meter of described severity of injuries Calculation method is:

${\mathrm{sp}}_s=\underset{p=1}{\overset{p}{\σ}}{f}_{ps}\×w_p$

Wherein, sp_sFor the severity of injuries index at preset location s, f_psFor the frequency of accident pattern p at the s of place, w_p For dead, severely injured and slight wound weight coefficient, p is the species number of the accident pattern considering.

The above-mentioned road accident Multiple trauma discrimination method based on spatial autocorrelation, wherein, the described spatial statisticses of described calculating Before the distribution results of standardization z value of the local statistic of unit, also include:

Calculate local gi statistic；

Calculate expectation e (gi) and variance var (gi) of local gi statistic；

Calculate standardization z (i) value of described local statistic gi.

The above-mentioned road accident Multiple trauma discrimination method based on spatial autocorrelation, wherein, described calculating local gi statistic Method be:

$gi\left(d\right)=\frac{{\mathrm{\σ}}_j{w}_{ij\left(d\right){\mathrm{sp}}_j}}{{\mathrm{\σ}}_j{\mathrm{sp}}_j};$

The computational methods of the expectation e (gi) of described local gi statistic are:

E (gi)=w/n；

The computational methods of variance var (gi) of described local gi statistic are:

$var\left(gi\right)=\frac{s^2}{\stackrel{-}{sp}}\×\frac{w_i(n-w_i)}{n-1};$

The method of the standardization z value of the described local statistic gi of described calculating is:

$z\left(i\right)=\frac{gi\left(d\right)-e\left(gi\right)}{\sqrt{var\left(g_{i\left(d\right)}\right)}};$

Wherein,sp_jFor the accident at space cell j The order of severity,For space severity of injuries meansigma methodss.

Correspondingly, the present invention also provides a kind of road accident Multiple trauma identification device based on spatial autocorrelation, comprising:

Model representation unit, for using the accident point to be detected in nonparametric probability model representation region to be measured Two dimensional Distribution；

Optimal bandwidth determining unit, for determining the optimal bandwidth h of described Density Estimator model₀；

Construction unit, for building spatial statistical units using " face-arc " network topology model based on geosystem；

Judging unit, for judging whether the incident attributes on described spatial statistical units assume spatial autocorrelation；

Computing unit, described for when the incident attributes on described spatial statistical units assume spatial autocorrelation, calculating The distribution results of standardization z (i) value of the local statistic of spatial statistical units；

Scattergram signal generating unit, for the distribution results generation accident of standardization z (i) value according to described local statistic The scattergram of Multiple trauma.

Due to adopting technique scheme, the invention has the following beneficial effects:

1. the hazardous location identification method and device based on spatial autocorrelation that the present invention provides, from the ground of distribution of traffic accident Reason angle set out, by between path space unit close on relation premised on, in conjunction with the incident attributes value on path space unit Describe its Characteristics of Geographical Distribution, carry out the discriminating of Accident Area using Density Estimator and spatial autocorrelation method, overcome Multiple trauma discrimination process in prior art accident is using the analysis method of classical mathematical statisticss, so that data visualization The poor defect of ability to express, the present invention is based on locus and geographical location information, makes full use of spatial analysis functions, to latent Lay the foundation in the excavation of cause of accident.

2. the hazardous location identification method and device based on spatial autocorrelation that the present invention provides, accident point to be measured is pressed It is the accident point of the accident point in the region of intersection and section region according to region division, adopt not in conjunction with the space characteristics of the two Same method builds the spatial relationship in region to be measured and accident point to be measured, then calculates severity of injuries evaluation number, the program Overcome in prior art and section region and crossing region evaluated respectively, and do not consider therebetween interactional lack Fall into.

The technical program from China's road traffic operation and management status, for path space build scientific and reasonable, Effectively convenient, the strong applicability and urban road hazardous location identification scheme of the accident origin cause of formation can be analyzed.

Brief description

Fig. 1 be one of road accident Multiple trauma discrimination method based on spatial autocorrelation in the embodiment of the present invention 1 concrete The flow chart of example；

Fig. 2 is the structural representation of preferably spatial statistical units in the embodiment of the present invention 1；

Fig. 3 a is the distribution results of the z value of the local statistic of accident number in the embodiment of the present invention 1 based on space cell Figure；

Fig. 3 b is dividing of the z value of the local statistic of severity of injuries in the embodiment of the present invention 1 based on space cell Cloth result figure；

Fig. 4 is the modular structure of the road accident Multiple trauma identification device in the embodiment of the present invention 2 based on spatial autocorrelation Figure.

Specific embodiment

Below in conjunction with shown in accompanying drawing, the present invention is further illustrated.

A kind of road accident Multiple trauma discrimination method based on spatial autocorrelation of the present invention, as shown in figure 1, include following walking Rapid:

Step s1, the two dimension using the accident point to be detected in nonparametric probability model representation region to be measured is divided Cloth, selectes the accident point to be measured in region to be measured, the Two dimensional Distribution of this accident point to be measured to be represented using nuclear density model, by Do not utilize the priori of the relevant distribution of accident point to be measured, the distribution to accident point to be measured in nonparametric probability method Add any it is assumed that from accident point in itself to be measured study accident point to be measured distribution characteristicss, therefore, the method application Field is paid attention to by height；

Step s2, determines the optimal bandwidth h of described Density Estimator model₀, by two of the accident point to be measured in step s1 Dimension distribution is represented with a kernel density function, calculates the integration of this kernel density function and real density according to the property of kernel density function Corresponding window width, as optimal bandwidth when mean square error is minimum；

Step s3, builds spatial statistical units using " face-arc " network topology model based on geosystem, introduce " face- Radian " topological model, to road network struction model, region to be measured is divided into intersection region and is located at two friendships of arbitrary neighborhood Section region between the domain of cross road mouth region, for intersection region, the then different qualities difference according to two regions to be measured Set up the spatial relationship of accident point to be measured in intersection and intersection and the thing to be measured in section region and section region Therefore the spatial relationship of point, explained with a specific embodiment as follows:

Structure shown in Figure 2, be 2316 of Urumqi City 2007 to 2011 put on record casualty data according to 2468 spatial statistical units that " face-arc " model partition is, in figure taking some sections of spatial statistical units as a example, by " face-arc " Topological model layer and casualty data layer overlay analysis, wherein, the data based on " face-arc " topological model layer is that traffic police in handling cases exists On google map, positioning obtains its longitude and latitude import arcgis system database the later stage, therefore, during use, directly from Obtain in arcgis system database, and the data (including but are not limited to the number of casualties, accident pattern etc.) of casualty data layer It is to be provided by traffic control department, is comprehensively analyzed, obtain a class region and b class region, wherein, a class region is crossroad mouth region Domain, b class region is section region, then occur the accident in a class region to be crossing accident, the thing in b class region occurs Therefore be section accident.

Step s4, judges whether the incident attributes on described spatial statistical units assume spatial autocorrelation, spatial autocorrelation Judgement need to judge whether moran index i is more than zero, and the standardization z value of expectation e (i) of moran index and variance construction Whether it is more than a preset value, when the two all meets, represent to be in spatial autocorrelation, otherwise, be not in spatial autocorrelation.With regard to described " preset value ", as shown in Figure 3 a and Figure 3 b shows, wherein z partly can be marked in figure more than preset value, and also can enter One step is according to z value distinguishing mark significance level.For example can using the corresponding part of the first color mark 1.65 < z < 1.96 simultaneously It is judged to three-level Accident Area；1.96 < z < 2.58 correspond to α=0.05 significance levels, show this unit be one relatively The space clustering of high level, it is possible to use the second color is marked to appropriate section and is judged to second grade accident Multiple trauma；Space Unit standardization z value > 2.58 correspond to α=0.01 significance level, show that this unit is the space clustering of an extremely high value, Using the 3rd color, appropriate section can be marked and be judged to one-level Accident Area, other space cells represent not to be deposited In obvious space clustering, it is possible to use other colors are marked.

Step s5, if the incident attributes on described spatial statistical units assume spatial autocorrelation, calculates described space system The distribution results of standardization z (i) value of local statistic of meter unit, refer to shown in Fig. 3 a, 3b, are to choose in the present embodiment Standardization z (i) value of the local statistic of accident number of 2468 spatial statistical units of Urumqi City distribution results, Wherein, Fig. 3 a is z (i) the Distribution value result of the local statistic based on space cell accident number, and Fig. 3 b is based on space cell Z (i) the Distribution value result of the local statistic of severity of injuries.

Step s6, the distribution results of standardization z (i) value according to described local statistic generate the distribution of Accident Area Figure.

This scattergram is the scattergram that the standardization z value according to the local statistic generating in above-mentioned steps s5 generates, tool The form of body scattergram is carried out selecting to set by client according to demand, and meanwhile, software is shown that based on which kind of incident attributes, example The scattergram of standardization z (i) value of the local statistic of accident number is also based on the local statistic of severity of injuries in this way Standardization z (i) value scattergram, or the combination of the two or multiple attribute, user is selected as needed.

The above-mentioned hazardous location identification method based on spatial autocorrelation, solves and cannot differentiate in prior art to include A difficult problem for Accident Area in the urban road of crossing, from the geographical angle of distribution of traffic accident, with path space unit Between close on relation premised on, in conjunction with the incident attributes value on path space unit, its Characteristics of Geographical Distribution is described, using core Density estimation and spatial autocorrelation method carry out the discriminating of Accident Area, overcome Accident Area mirror of the prior art Other process is using the analysis method of classical mathematical statisticss, so that the poor defect of data Visualization ability, the present invention Based on locus and geographical location information, make full use of spatial analysis functions, the excavation to latent defect reason lays the foundation.

Preferably, the kernel density function in step s2 is:

$\hat{u}\left(x\right)=\frac{1}{{\mathrm{nh}}^d}{\mathrm{\&sigma;}}_{i=1}^{n}k\left(\frac{l-l_i}{h}\right);$

Wherein, k () is kernel function, and h is window width, and n is the number of the accident point to be measured in region to be measured, and d is the dimension of data Number, (l-l_i) for accident point to be measured to accident point l_iDistance.

Preferably, determine the optimal bandwidth of described kernel density function in step s2, also include:

Step s21, obtains real density f (x) in described region to be measured for the described accident point to be measured；

Step s22 is right when the described kernel density function of acquisition is minimum with the integrated square error of described real density f (x) The window width h answering₀, as optimal bandwidth.

On this basis, further, kernel density function and the integrated square error computing formula of real density f (x) are:

WhereinFor kernel density function,For integration all Square error.

The present embodiment is divided into basic road using 1000m for unit for road, then close according to core for primary and secondary main line The optimal window width of degree function determining, total accident sample point totally 1661 on the primary and secondary main line of survey region in the present embodiment Individual, from the density of gaussian kernel function estimation plane space accident point, obtain optimal bandwidth h₀For 115.2m.

Preferably, in step s3, spatial statisticses list is built using " face-arc " network topology model based on geosystem Unit, comprises the steps:

Step s31, region to be measured is divided into intersection region and section region；

Step s32, for the accident point in the region of intersection, sets up crossroad by the operational rule in " face " and " point " Mouthful with the spatial relationship of accident；

Step s33, for the accident point in section region, overlaps, by spatial data, the space that road and accident are set up in computing Relation, and overlapped before budget sets up spatial relationship by spatial data described, in GIS-Geographic Information System (geographic Information system, abbreviation gis) in set the relief area of predetermined width for segmental arc.

Using the method, by the accident point in the section region in the urban road including crossing and intersection region Interior accident point is made a concrete analysis of according to its concrete condition, sets up spatial relationship and road and the accident of intersection and accident respectively Spatial relationship, by the method, according to the difference of accident generation area, set up the spatial relationship of accident and spot, comprehensive The relation closing analysis accident with spot carries out next step analysis.

Preferably, in step s4, whether the incident attributes on decision space statistic unit are in spatial autocorrelation, also include:

Step s41, calculates autocorrelation exponent i, expectation e (i) and expectation e (i) of incident attributes and the standard of variance construction Change z₁Value；What the autocorrelation exponent i of accident number reflected is the dependency of the incident attributes in region of Spatial Adjacency or spatial neighbor, Wherein, incident attributes can be but be not limited only to accident number, severity of injuries.

Step s42, judges whether the autocorrelation exponent i of described incident attributes is more than zero, and if so, statistic unit is then described The incident attributes of adjacent domain are in space positive correlation, continue step s43, otherwise, exit this operation, wherein, described auto-correlation The standardization z that the expectation of index is constructed with variance₁Value be more than preset value when, represent statistic unit on described Spatial Adjacency or sky Between the incident attributes in neighbouring region be in spatial autocorrelation.In addition, working as i < 0, represent that the accident of the adjacent domain of statistic unit belongs to Property negatively correlated in space, if i=0, represent that the incident attributes of the adjacent domain of statistic unit assume space randomness.

Step s43, the standardization z constructing with variance when the expectation of described autocorrelation exponent₁Value be more than preset value when, table Show between the incident attributes distribution pattern in the region of described Spatial Adjacency or spatial neighbor on statistic unit and random distribution pattern There is significant difference, and be in spatial autocorrelation.

Preferably, comprehensive analysis accident number and two incident attributes of severity of injuries, in the notable water of preferred α=0.01 Under flat, it is i by the moran index that above-mentioned formula is calculated respectively for accident number₁=0.24, z₁=12.01；For thing Therefore the moran index of the order of severity is i₂=0.15, z₂=7.6 it can be seen that the expectation of two autocorrelation exponents and variance structure The standardization z value made (includes z₁And z₂) be all higher than default 2.58, then show that the incident attributes on this spatial statistical units are in all sky Between auto-correlation.

By the method, effectively judge whether the incident attributes on spatial statistical units assume sky auto-correlation, different The spatial relationship of accident and spot is set up according to different principles in region, but, this step is judged different using identity principle Whether the incident attributes in region, all in spatial autocorrelation, have accomplished to analyze the accident point of zones of different respectively and unitary analysiss Purpose, solves each class accident point of isolated analysis and leads to not effectively set up the connection between the accident point in different incident regions The defect of system.

On this basis, further, the computational methods of moran index i are:

$i=\frac{{\mathrm{n\&sigma;\&sigma;w}}_{ij\left({\mathrm{sp}}_{i-}\stackrel{\&overbar;}{s}p\right)\left({\mathrm{sp}}_{j-}\stackrel{\&overbar;}{s}p\right)}}{\left({\mathrm{\&sigma;\&sigma;w}}_{ij}\right)\left(\mathrm{\&sigma;}{\left({\mathrm{sp}}_{i-}\stackrel{-}{sp}\right)}^2\right)};$

Wherein, w_ijFor the Spatial weight matrix of position relationship between space cell i and j, sp_i, sp_jFor space cell i and j The accident number at place or severity of injuries；For the meansigma methodss of accident number or severity of injuries, n is space cell sum.

Preferably, in above-mentioned steps s4, before judging whether the incident attributes on spatial statistical units assume spatial autocorrelation, Also include:

Step s ' 4, determines that the incident attributes of the present embodiment are severity of injuries and/or accident number.

On this basis, further, the method calculating severity of injuries is:

${\mathrm{sp}}_s=\underset{p=1}{\overset{p}{\&sigma;}}{f}_{ps}\&times;w_p$

Wherein, sp_sFor the severity of injuries index at preset location s, f_psFor the frequency of accident pattern p at the s of place, w_p For dead, severely injured and slight wound weight coefficient, p is the species number of the accident pattern considering, in the present embodiment, w_pFor dead, Severely injured and slight wound weight coefficient is respectively 9.5,9.5,3.5.

Before calculating severity of injuries evaluation number, give weight to accident first, then calculate severity of injuries and comment Valency index, in the present embodiment, the average accident number of space cell is 0.66, and average severity of injuries is 3.67.

Preferably, step s5, calculate spatial statistical units standardization z (i) value of local statistic distribution results it Before, also include:

Step s ' 51, calculates local gi statistic；

Step s ' 52, calculates expectation e (gi) and variance var (gi) of local gi statistic；

Step s ' 53, calculates standardization z (i) value of described local statistic gi.

On this basis, further, the method calculating local gi statistic is:

$gi\left(d\right)=\frac{{\mathrm{\&sigma;}}_j{w}_{ij\left(d\right){\mathrm{sp}}_j}}{{\mathrm{\&sigma;}}_j{\mathrm{sp}}_j};$

The computational methods of the expectation e (gi) of local gi statistic are:

E (gi)=w/n；

The computational methods of variance var (gi) of local gi statistic are:

$var\left(gi\right)=\frac{s^2}{\stackrel{-}{sp}}\&times;\frac{w_i(n-w_i)}{n-1};$

The method calculating standardization z (i) value of described local statistic gi is:

$z\left(i\right)=\frac{gi\left(d\right)-e\left(gi\right)}{\sqrt{var\left(g_{i\left(d\right)}\right)}};$

Wherein,sp_jFor the accident at space cell j The order of severity,For space severity of injuries meansigma methodss.

Preferably, when α=0.01, as z (i) > 2.58 when, represent that this unit is an Accident Area, referred to as one-level Accident Area；When α=0.05,1.96 < α=0.01 < when 2.58, the Frequent Accidents probability of this unit is less than this one-level accident Multiple trauma, setting this unit is a second grade accident Multiple trauma, wherein, i represent based on incident attributes type.

Being calculated for the Accident Area that accident number identifies by above-mentioned formula is at 53, for the serious journey of accident Spending the Accident Area identifying is at 47, and the Accident Area of the repetition being identified based on both attributes is at 28, then This 28 accident points are Accident Area, need to be improved.

In sum, the present invention is from the geographical angle of distribution of traffic accident, to close on relation between path space unit Premised on, in conjunction with the incident attributes value on path space unit, its Characteristics of Geographical Distribution is described, using Density Estimator and sky Between autocorrelation method carry out the discriminating of Accident Area, overcome Multiple trauma discrimination process in prior art accident using classical The analysis method of mathematical statisticss so that the poor defect of data Visualization ability, the present invention be based on locus and Geographical location information, makes full use of spatial analysis functions, and the excavation to latent defect reason lays the foundation；Meanwhile, by thing to be measured Therefore put the accident point for the accident point in the region of intersection and section region according to region division, in conjunction with the space characteristics of the two Build the spatial relationship in region to be measured and accident point to be measured using different methods, then calculate severity of injuries evaluation number, The program overcomes in prior art and evaluates section region and crossing region respectively, and does not consider mutual shadow therebetween The defect rung.

A kind of road accident Multiple trauma identification device based on spatial autocorrelation realizing said method, as shown in figure 4, bag Include:

Model representation unit 41, for using the accident to be detected in nonparametric probability model representation region to be measured The Two dimensional Distribution of point；

Optimal bandwidth determining unit 42, for determining the optimal bandwidth h of described Density Estimator model₀；

Construction unit 43, for building spatial statistical units using " face-arc " network topology model based on geosystem；

Judging unit 44, for judging whether the incident attributes on described spatial statistical units assume spatial autocorrelation；

Computing unit 45, for when the incident attributes on described spatial statistical units assume spatial autocorrelation, calculating institute State the distribution results of the standardization z value of the local statistic of spatial statistical units；

Scattergram signal generating unit 46, for the distribution results generation accident of the standardization z value according to described local statistic The scattergram of Multiple trauma.

The above-mentioned description to embodiment is to be understood that for ease of those skilled in the art and apply this Bright.Person skilled in the art obviously easily can make various modifications to these embodiments, and described herein General Principle is applied in other embodiment without through performing creative labour.Therefore, the invention is not restricted to above-described embodiment, , according to the announcement of the present invention, the improvement made for the present invention and modification all should be in the protections of the present invention for those skilled in the art Within the scope of.

Claims (10)

1. a kind of road accident Multiple trauma discrimination method based on spatial autocorrelation is it is characterised in that comprise the steps:

Two dimensional Distribution using the accident point to be detected in nonparametric probability model representation region to be measured；

Determine the optimal bandwidth h of described Density Estimator model₀；

Spatial statistical units are built using " face-arc " network topology model based on geosystem；

Judge whether the incident attributes on described spatial statistical units assume spatial autocorrelation；

If the incident attributes on described spatial statistical units assume spatial autocorrelation, calculate the local of described spatial statistical units The distribution results of standardization z (i) value of statistic；

The distribution results of standardization z (i) value according to described local statistic generate the scattergram of Accident Area.

2. according to claim 1 the road accident Multiple trauma discrimination method based on spatial autocorrelation it is characterised in that described The kernel density function of Density Estimator model is:

$\hat{u}\left(x\right)=\frac{1}{{\mathrm{nh}}^d}{\&sigma;}_{i=1}^{n}k\left(\frac{l-l_i}{h}\right);$

Wherein, k () is kernel function, and h is window width, and n is the number of the accident point to be measured in region to be measured, and d is the dimension of data, (l-l_i) for accident point to be measured to accident point l_iDistance.

3. according to claim 1 the road accident Multiple trauma discrimination method based on spatial autocorrelation it is characterised in that described Determine the optimal bandwidth of described kernel density function, comprising:

Obtain real density f (x) in described region to be measured for the described accident point to be measured；

Obtain described kernel density function and the corresponding window width h during integrated square error minimum of described real density f (x)₀, that is, For optimal bandwidth.

4. according to claim 3 the road accident Multiple trauma discrimination method based on spatial autocorrelation it is characterised in that described Kernel density function with the integrated square error computing formula of described real density f (x) is:

WhereinFor described kernel density function,Amass for described Divide mean square error.

5. according to claim 1 the road accident Multiple trauma discrimination method based on spatial autocorrelation it is characterised in that described Spatial statistical units are built using " face-arc " network topology model based on geosystem, comprising:

Region to be measured is divided into intersection region and section region；

Accident point in the region of intersection, sets up the space pass of intersection and accident by the operational rule in " face " and " point " System；

The accident point in section region, overlaps, by spatial data, the spatial relationship that road and accident are set up in computing, and described Budget is overlapped by spatial data and sets up before spatial relationship, GIS-Geographic Information System sets for segmental arc the relief area of predetermined width.

6. according to claim 1 the road accident Multiple trauma discrimination method based on spatial autocorrelation it is characterised in that described Judge whether the incident attributes on described spatial statistical units assume spatial autocorrelation, comprising:

Step s41, calculates autocorrelation exponent i, expectation e (i) and expectation e (i) of incident attributes and the standardization z of variance construction₁ Value；

Step s42, judges whether the autocorrelation exponent i of described incident attributes is more than zero, and if so, the neighbouring of statistic unit be then described The incident attributes in region are in space positive correlation, continue step s43, otherwise, exit this operation；

Step s43, the standardization z that the expectation of described autocorrelation exponent is constructed with variance₁Value be more than preset value when, represent statistics On unit, the incident attributes in the region of described Spatial Adjacency or spatial neighbor are in spatial autocorrelation；

Preferably, the computational methods of described autocorrelation exponent i are:

$i=\frac{n\&sigma;\&sigma;w_{ij\left({\mathrm{sp}}_{i-}\stackrel{\&overbar;}{s}p\right)\left({\mathrm{sp}}_{j-}\stackrel{\&overbar;}{s}p\right)}}{(\&sigma;\&sigma;w_{ij})(\&sigma;{\left({\mathrm{sp}}_{i-}\stackrel{\&overbar;}{s}p\right)}^2)};$

Wherein, w_ijFor the Spatial weight matrix of position relationship between space cell i and j, sp_i, sp_jAt space cell i and j Accident number or severity of injuries；For the meansigma methodss of accident number or severity of injuries, n is space cell sum.

7. according to claim 1 the road accident Multiple trauma discrimination method based on spatial autocorrelation it is characterised in that in institute State judge described spatial statistical units on incident attributes whether assume spatial autocorrelation before, also include:

Determine that incident attributes are severity of injuries and/or accident number；

Preferably, the computational methods of described severity of injuries are:

${\mathrm{sp}}_s=\underset{p=1}{\overset{p}{\&sigma;}}{f}_{ps}\&times;w_p$

Wherein, sp_sFor the severity of injuries index at preset location s, f_psFor the frequency of accident pattern p at the s of place, w_pFor dead Die, the weight coefficient of severely injured and slight wound, p is the species number of the accident pattern considering.

8. according to claim 1 the road accident Multiple trauma discrimination method based on spatial autocorrelation it is characterised in that described Before the distribution results of standardization z value of the local statistic calculating described spatial statistical units, also include:

Calculate local gi statistic；

Calculate expectation e (gi) and variance var (gi) of local gi statistic；

Calculate standardization z (i) value of described local statistic gi.

9. according to claim 8 the road accident Multiple trauma discrimination method based on spatial autocorrelation it is characterised in that described The method calculating local gi statistic is:

$gi\left(d\right)=\frac{{\&sigma;}_j{w}_{ij\left(d\right){\mathrm{sp}}_j}}{{\mathrm{\&sigma;}}_j{\mathrm{sp}}_j};$

The computational methods of the expectation e (gi) of described local gi statistic are:

E (gi)=w/n；

The computational methods of variance var (gi) of described local gi statistic are:

$var\left(gi\right)=\frac{s^2}{\stackrel{\&overbar;}{s}p}\&times;\frac{w_i(n-w_i)}{n-1};$

The method of the standardization z value of the described local statistic gi of described calculating is:

$z\left(i\right)=\frac{gi\left(d\right)-e\left(gi\right)}{\sqrt{var\left(g_{i\left(d\right)}\right)}};$

Wherein,sp_jSerious for the accident at space cell j Degree,For space severity of injuries meansigma methodss.

10. a kind of road accident Multiple trauma based on spatial autocorrelation realizing arbitrary methods described in claim 1 to 9 differentiates Device is it is characterised in that include:

Model representation unit, for using two of the accident point to be detected in nonparametric probability model representation region to be measured Dimension distribution；

Optimal bandwidth determining unit, for determining the optimal bandwidth h of described Density Estimator model₀；

Construction unit, for building spatial statistical units using " face-arc " network topology model based on geosystem；

Judging unit, for judging whether the incident attributes on described spatial statistical units assume spatial autocorrelation；

Computing unit, for when the incident attributes on described spatial statistical units assume spatial autocorrelation, calculating described space The distribution results of the standardization z value of the local statistic of statistic unit；

Scattergram signal generating unit, the distribution results for the standardization z value according to described local statistic generate Accident Area Scattergram.