Specific implementation mode
The present embodiment realizes a kind of pedestrian traffic accident method for early warning suitable for vehicle-mounted short distance communication network, purpose
It is to receive surrounding vehicles by pedestrian's mobile device (such as by multi mode terminal with vehicle-mounted short-range communication module) periodically to send out
That send includes the message (hereinafter referred to as heartbeat message) of geographical position coordinates and vehicle operating information etc., obtains surrounding vehicles with this
The information such as distance, speed, acceleration, traveling and the turn direction of vehicle, meanwhile, receive the traffic that trackside node periodic broadcasting is sent
Then the information such as signal lamp state utilize Bayesian network, to that may cause multiple items of information of traffic accident, by calculating shellfish
Joint probability between this network variable of leaf calculates the probability of accident generation, when probability is more than threshold value, to pedestrian in time into
Row early warning achievees the purpose that reduce traffic accident.
Bayesian network is a kind of method carrying out probability inference in artificial intelligence field, can utilize the part in model
Condition dependence carries out two-way uncertain inference, be applied to analysis and judge etc..The traffic accident of Pedestrians and vehicles is by multiple
The influence of factor, prediction is complex, and there is complicated contacts between predictive factors, is that realization is changeable using Bayesian network
A kind of effective ways of amount analysis and evaluation.
The present embodiment is made of five parts.
First part, vehicle running state classification.
Vehicle occurs mainly in the vehicle around pedestrian to the threat of pedestrian, include in pedestrian's sight and sight outside vehicle
, vehicle, hot-short especially outside pedestrian's sight, the vehicle intersected with pedestrian direction easily cause traffic accident.It is comprehensive
The transport condition of vehicle is divided into three classes by above-mentioned analysis, the present embodiment.
1.1, first kind vehicle-state is front and back car state.Main includes the vehicle in the same direction and reversed with pedestrian, the first kind
Vehicle-state is again related with the relative position of people's vehicle, vehicle heading, because traffic intersection is accident-prone road section, at crossing
It is also related with the state of signal lamp.For people's vehicle relative distance, main same car speed, people's vehicle distance and pedestrian walking side
To related.
1.2, the second class vehicle-state is left and right vehicle.Mainly pedestrians travel's road or walking to belt road be face
To ridden in left or right direction vehicle.It is similar with first kind vehicle-state, relative position of the second class vehicle-state with people's vehicle, vehicle row
It is related to sail directional correlation, the state of signal lamp.
The framework determination of second part, Bayesian network
Bayesian network includes two parts:(1) the bayesian network structure figure of directed acyclic, node on behalf variable are oriented
Arc represents interrelated;(2) conditional probability table between node and node indicates the relationship of relationship between node.
Bayesian network is with N={ < V, E >, P }, wherein D indicates the acyclic figure (DAG) of Bayes, X=(X1,
X2,...,Xn) it is node set, joint probability distributionWherein, Parent
(xi) indicate xiThe father node of node.If obtaining the prior probability and node of each interdependent node from the statistical result of existing information
Between conditional probability, using p (X) can reasoning other nodes probability.Bayesian network (BN) inference method is, given
When to the actual value of variables collection E, the posteriority conditional probability distribution P (Q | E) of variables collection Q is calculated.
2.1, the Bayesian network framework of accident probability determines.In the present embodiment, the front and back vehicle situation of accident probability selection, a left side
Right vehicle situation is predicted as parameter.
2.2, various types of vehicles situation selects relative distance, is predicted close to situation as parameter.
2.3, the relative distance of various types of vehicles selects car speed, people's vehicle distance as parameter prediction.
2.4, various types of vehicles close to situation pedestrian direction, signal(l)ing condition as parameter prediction.
Complete accident forecast Bayesian network framework is as shown in Figure 2.
Conditional probability determination between Part III, Bayes's node and network training.
3.1, it is directed to each node, each node is obtained with the condition between father node by expertise and historical data
Probability.By taking the universal nodes that Fig. 3 is indicated as an example, the conditional probability of node A and father node A1 and A2 according to expertise or are gone through
The determination of history data, as shown in table 1.
Table 1
Wherein, { A1-1, A1-2 ... ..., A1-m } indicates m different conditions of father node variables A 1, { A2-1, A2-
2 ... ..., A2-n } indicate n different conditions of father node variables A 2, { A1, A2 } indicates a different groups of m × n of two variables
It closes.{ s-1, s-2 ... ..., s-k } indicates that the k different states of node A, Pi-j indicate the A sections of corresponding certain { A1, A2 } combination
Point is (s-j) shape probability of state, is hadThe item of each node and its father node can be obtained successively using the method
Part probability.With (relative distance A1) node and two father node (two vehicle speed A11), (people's spacing in Fig. 2 in the present embodiment
From A12) for, A11 takes 1 (quick)/0 (common) respectively, and A12 takes 1 (close)/0 (common) respectively, and A1 takes respectively 1 (has danger
Danger)/0 (no danger), then according to expertise or historical data, it may be determined that { A11, A12 }={ 1,1 }, { A1=1's } is general
Rate is 86%, and the probability of { A1=0 } is 14%.The probability of { A11, A12 }={ 1,0 }, { A1=1 } is 36%, and { A1=0's } is general
Rate is 64%.The probability of { A11, A12 }={ 0,1 }, { A1=1 } is 22%, and the probability of { A1=0 } is 78%.{ A11, A12 }=
The probability of { 0,0 }, { A1=1 } is 3%, and the probability of { A1=0 } is 97%.
3.2, Bayesian network is trained.Using the historical data of the traffic accident in this section, as prior probability, in conjunction with each
Conditional probability between a node obtains the joint probability of each variable, carries out the training of Bayesian network, according to same history number
According to the degree that is consistent, adjust each conditional probability so that Bayesian network is more efficient.
3.3, the parameter of Bayesian network is issued by system by the trackside node in this section, pedestrian's mobile terminal
Into within the scope of the trackside coverage in this section, Bayes's parameter is downloaded from the trackside node in this section, so that Bayes
Prediction meets the actual conditions in this section.
The acquisition and calculating of Part IV, vehicle-state and pedestrian's state.
4.1, the acquisition of vehicle-state.In V2X networks, vehicle periodicity sending includes that the heartbeat of geographical location information is believed
The mobile terminal of breath, pedestrian can obtain the location information of surrounding vehicles by the heartbeat message of reception vehicle, believe by geography
The computational methods of breath system, can calculate the distance, speed, travel direction of vehicle, so can calculate prediction people's vehicle between
Relative distance.
4.2, the acquisition of pedestrian's state.The geographical location device (GPS or the Big Dipper) being equipped with by pedestrian's terminal, can obtain
The geographical location of pedestrian can calculate the direction of travel of pedestrian, speed etc..
4.3, the acquisition of signal lamp state.In V2X, trackside node is generally disposed in road cross, periodically by signal lamp shape
State is to surrounding broadcast.The mobile terminal of pedestrian, the information by receiving trackside node can obtain the signal of crossing all directions
The state of lamp.
Part V, traffic accident prediction and alarm.
5.1, the mobile terminal of pedestrian is derived after obtaining relevant information using Bayesian network, to finally obtain
Obtain the probability that traffic accident occurs under current environment.In Bayesian network, it is known that the item between the probability and node of father node
Part probability, you can to obtain final conditional probability, be equal to the product of all conditions probability distribution, i.e. joint probability distribution.Have
Wherein, Parent (xi) indicate xiThe father node of node.By taking three nodes of Fig. 3 as an example, it is known that father node
Save A1、A2The probability and node A of point are the same as father node A1、A2Conditional probability, then calculate P1=P (AA1A2)=P
(A1)P(A2)P(A|A1A2), Further obtainUsing same method, you can be pushed away step by step from the top to Bayesian network
It arrives, it is final to obtain the lowermost traffic accident probability of happening.
If 5.3, the probability of traffic accident is more than the threshold value of default, audible and visual alarm is carried out, and prompt pedestrian
Maximum probability factor in Bayesian network.
Through this embodiment, can in vehicle-mounted short-range communication network, the mobile terminal of pedestrian passes through receive around it is vehicle-mounted
The heartbeat message of the heartbeat message and trackside node broadcasts of node periodic broadcasting obtains the positions of surrounding vehicles, speed, acceleration
Information is handled much information by Bayesian network, to predict in time the traffic accident that may occur.
The above is presently preferred embodiments of the present invention and its technical principle used, for those skilled in the art
For, without departing from the spirit and scope of the present invention, any equivalent change based on the basis of technical solution of the present invention
Change, simple replacement etc. is obvious changes, all fall within the protection scope of the present invention.