Arc length ratio sigma_jAnd dividing the numerical value into three groups of data by using a hierarchical clustering algorithm as a characteristic value, and determining an abnormal driving track, an abnormal behavior track and a normal vehicle track according to the data volume and the dispersion degree epsilon of each group of data. The data with the minimum data volume and the ratio of the number of the groups to the total number of the groups smaller than the abnormal threshold value is defaulted as an abnormal track, and the data with the larger dispersion degree epsilon is defaulted as an abnormal behavior track;

and 2-1-3, performing hierarchical clustering on the normal vehicle tracks in the step 2-1-2 again based on a DTW (delay tolerant shift) algorithm, judging and dividing the normal vehicle tracks and the outlier tracks based on the track number and the dispersion degree epsilon of each cluster after clustering, and extracting the normal vehicle tracks.

Further, the step 2-2 specifically comprises the following sub-steps:

step 2-2-1, normal vehicle track TR extracted by screening_jLower trace point p_iRealizing the smoothness of the vehicle track, equally selecting and determining new characteristic points p 'of the vehicle track according to the smoothed vehicle track points'_i；

Step 2-2-2, based on all vehicle trajectories TR_jAnd novel characteristic point p'_iDetermining a main curve of vehicle operation; specifically, feature points p 'are selected based on each vehicle track'_iCorresponding the main curves one by one, determining the central point of the stage i in the same stage i to obtain the main curve TR^midNamely:

in the formula, TR^midRepresenting a vehicle trajectory master curve; p_i^midRepresenting the central point of the i-th stage track characteristic point, wherein i is more than or equal to 1 and less than or equal to n, and n represents the number of the selected characteristic points; p'_jiRepresenting the new characteristic points of the ith stage track of the jth vehicle track, wherein j is more than or equal to 1 and less than or equal to N, and N represents the total number of the vehicle tracks;

and simultaneously determining outliers of each stage based on the numerical distribution of the distance between each stage and the central point, namely:

of formula (II b), dist (p'_ji,P_i^mid) Representing the distance between the characteristic point of j tracks i and the central point of the stage i;

analyzing the distance values of all track points in the stage i, and defaulting the track stage points corresponding to the values which are greater than normal distribution (mu +2 sigma) to be outliers through normal distribution analysis;

step 2-2-3, according to the main curve TR^midAnd a phase point P on the main curve_i^midDetermining envelope point Q of each stage_iThen, two envelope lines are obtained smoothly, and a reasonable driving range of the vehicle track is determined;

further, the step 2-2-3 specifically comprises the following sub-steps:

step 2-2-3-1, according to stage point P on main curve_i^midDetermining vehicle track points p 'in the same stage'_jiWith main curve TR^midIf the point of the track closest to the main curve is not on the main curve TR^midIf so, indicating that the vehicle track points at the stage deviate, and rejecting the vehicle track points, otherwise, turning to the next step;

step 2-2-3-2, track points p 'on each vehicle track which is not deviated in the same stage'_jiWith main curve TR^midDistance dist (p'_ji,TR^mid) Sorting, selecting 80-90% median value as radius distance of envelope curve corresponding to the main curve at the stage, and recording as r_i；

Step 2-2-3-3, based on the stage point P on the main curve_i^midDetermining envelope point coordinates Q at the same stage_iIs recorded as (x'_i,y′_i) And in particular,

y′_i＝k_ix′_i+b_i

in formula (II), x'_iAnd y'_iPoint coordinates representing the envelope at stage i, denoted as Q_i，

And

denotes the main curve coordinate in phase i, denoted P_i^mid；k_iRepresents the slope of the envelope in phase i, b_iRepresenting the intercept of the intersection line of the envelope curve and the main curve in the i stage;

step 2-2-3-4, judging the position relation between the envelope coordinate and the main curve according to the envelope coordinate solved in the previous step; in particular, the main curve point P according to this phase_i^midAnd the main curve point of the next stage

An envelope threshold ξ is obtained, namely:

in formula (II), x'_iAnd y'_iPoint coordinates representing the envelope at stage i;

and

representing the main curve coordinates on stage i;

x 'if ξ < 0'_iAnd y'_iThe point of the envelope curve is on the left side of the main curve, and ξ > 0 represents x'_iAnd y'_iThe envelope curve point is on the right side of the main curve, otherwise, the envelope curve point is on a straight line intersecting the main curve;

step 2-2-3-5, coordinate Q of envelope curve point of all stage points_iIntegrating, smoothing envelope track to obtain new envelope characteristic point q_s(i) To thereby determine the envelope trajectory TR^qAnd the range formed by the two lines is the reasonable driving range of the vehicle track in the mode.

Further, thestep 3 includes the following sub-steps:

step 3-1, determining the running direction and track information of the vehicle according to the real-time monitoring video analysis;

step 3-2, comparing the vehicle track path obtained in the step 3-1 with the reasonable driving range of the vehicle track at the intersection obtained in the step 2; if the vehicle exceeds the driving range, early warning is carried out, and vehicle information is issued to an off-site law enforcement system;

and 3-3, if the ratio of the number of early warning vehicles in the running flow direction in unit time to the total number of passing vehicles is greater than the threshold value of the abnormal condition, early warning the flow direction of the intersection, and simultaneously associating the flow direction to a traffic situation monitoring system or a video monitoring system, so that the alarm of the abnormal condition of the intersection is realized, and the abnormal condition of the intersection is monitored.

The invention achieves the following beneficial effects: compared with the traditional GPS vehicle track clustering abnormal analysis mode, the method provided by the invention realizes extraction and analysis of the vehicle track at the intersection by a hierarchical clustering algorithm on the basis of historical video data, effectively identifies vehicle tracks such as abnormal vehicle tracks, abnormal behavior tracks and outlier tracks, and extracts correct vehicle tracks. Compared with the traditional crossing illegal abnormal early warning mode, the invention extracts and analyzes the correct vehicle track, determines the reasonable driving range of each flow direction vehicle of the crossing, has smaller and more reasonable track interval than the original crossing, effectively warns the abnormal illegal driving vehicle, improves the non-field law enforcement efficiency, reduces the workload of manual examination and verification, and provides a reference index for intelligent identification; meanwhile, the abnormal conditions of the intersection are judged based on the abnormality of a plurality of vehicles, so that remote linkage monitoring is realized, and the supervision efficiency of intersection traffic is improved.

Drawings

Fig. 1 is a schematic step diagram of an abnormal driving early warning method according to the present invention.

FIG. 2 is a schematic diagram of a vehicle track pattern partitioned by hierarchical clustering according to an embodiment of the present invention.

FIG. 3 is a schematic diagram of a single intersection vehicle trajectory extracted according to an embodiment of the present invention.

Detailed Description

The technical scheme of the invention is further explained in detail by combining the drawings in the specification.

step 1, determining each vehicle track of the intersection based on the intersection monitoring historical video, and classifying the vehicle track type modes through hierarchical clustering and intersection canalization.

Step 1-1, based on historical videos of intersection monitoring unit time periods, extracting original track points under different vehicle IDs, marking the original track points as P (f, x, y), wherein f represents a frame number, x and y represent coordinate values, further determining a single vehicle track of the different vehicle IDs through initial data cleaning, and marking the single vehicle track as TR (TR) { P | P }_iI is more than or equal to 1 and less than or equal to n, and n is the number of track points }; in general, the unit time period is 5 min.

Step 1-2, classifying the vehicle track type modes based on a hierarchical clustering algorithm and intersection channelized information. Specifically, the feature point p is performed on a single vehicle track_iAnd extracting, namely performing vehicle track hierarchical clustering through the Euclidean distance or the LCSS distance, eliminating abnormal vehicle tracks (such as vehicle running tracks of stopping at the intersection, reverse running and the like), and realizing vehicle track division according to intersection canalization characteristics. Generally, taking an intersection as an example, the vehicle trajectory types include 12 types of left turn, straight turn and right turn in the east, south, west and north directions.

And 2, extracting and analyzing the reasonable range of the vehicle track in each flow direction of the intersection respectively, and realizing the reasonable range management of the vehicles at the intersection.

And 2-1, extracting the single-mode type vehicle track TR, and analyzing all tracks under the type through a hierarchical clustering algorithm to realize normal vehicle track extraction.

Step 2-1-1, based on vehicle trajectory TR_jCharacteristic point p_iTrack similarity lambda and variance alpha of single vehicle track²And the arc length ratio sigma is used as a characteristic numerical value to solve. Wherein i is more than or equal to 1 and less than or equal to n, and n is the number of the track characteristic points; j is more than or equal to 1 and less than or equal to N, and N is the number of the vehicle tracks.

In particular, the vehicle trajectory TR is solved by the LCSS algorithm or DTW algorithm according to the vehicle trajectory type_jSimilarity lambda; based on vehicle trajectory TR_jCharacteristic point p of_iDetermines the vehicle track TR by the number of frames and the distance between the feature points_jAcceleration variance α of²(ii) a Based on vehicle trajectory TR_jDetermining the vehicle trajectory TR by the distance between the characteristic points_jThe arc length ratio σ.

Wherein, taking LCSS algorithm as an example, the vehicle track TR_jThe similarity λ formula is:

wherein TR is₁And TR₂Two tracks with the length of m and n are respectively arranged;

wherein p is_iAnd q is_jRespectively represent the coordinates of the characteristic points of the track,

in the same way

Wherein i is more than or equal to 1 and less than or equal to m, j is more than or equal to 1 and less than or equal to n,

indicating that the track is empty; dist (p)_i,q_j) Representing Euclidean distance of two coordinate points, wherein gamma is a similarity threshold value; LCSS (TR)₁,TR₂) Represents TR₁And TR₂The longest common subsequence length of the two tracks; d_LCSS(TR₁,TR₂) Is a track TR₁And TR₂The similarity distance of (2); min (len)_TR1,len_TR2) Representing the track TR₁Length and TR₂The smaller value of the length; d_LCSS(TR₁,TR_l) Represents TR₁And TR_lThe longest common subsequence similarity distance between, N represents the number of vehicle tracks; lambda [ alpha ]₁Represents TR₁Phase of vehicle trajectoryA similarity value.

Vehicle track TR_jAcceleration variance α of²Comprises the following steps:

in the formula, alpha_i+1Represents the acceleration of the characteristic point i +1, where f_iAnd f_i+1Which represents the number of frames,

denotes the Euclidean distance, p, between the feature point i and the feature point i +1_iAnd p_i+1The feature points are represented.

Vehicle track TR_jThe arc length ratio σ of the trajectory of (a) is:

in the formula, p₁、p_i、p_i+1、p_nAll represent vehicle trajectory TR_jAn inner feature point;

representing the Euclidean distance between the vehicle track characteristic point i and the characteristic point i + 1;

representing the Euclidean distance between the characteristic point n of the vehicle root track and the characteristic point 1; n represents the number of characteristic points in the vehicle track, and i is more than or equal to 1 and less than or equal to n.

Step 2-1-2, the vehicle track TR_jThe similarity of the two components is lambda,Acceleration variance α²Arc length ratio sigma_jAnd performing hierarchical clustering on the vehicle track as characteristic data, dividing the characteristic data into three types, namely abnormal driving track, abnormal behavior track and normal vehicle track, and further removing the abnormal driving track and the abnormal behavior track to realize normal vehicle track data extraction.

Specifically, each vehicle track TR is integrated_jSimilarity of (D) ("lambda")_jVariance of acceleration

Arc length ratio sigma_jAnd dividing the numerical value into three groups of data by using a hierarchical clustering algorithm as a characteristic value, and determining an abnormal driving track, an abnormal behavior track and a normal vehicle track according to the data volume and the dispersion degree epsilon of each group of data. The data set with the minimum data volume and the ratio of the data volume to the total number smaller than the abnormal threshold (generally 30%) is defaulted as an abnormal track, and the data set with the greater dispersion degree epsilon is defaulted as an abnormal behavior, wherein the dispersion degree epsilon is as follows:

in the formula, n' is the total number of all vehicle track characteristic points in the divided group; p is a radical of_iIn order to be a characteristic point, the method comprises the following steps of,

the cluster center point is the group center of the divided group data;

represents p_iPoint to cluster center point distance.

And 2-1-3, performing hierarchical clustering on the normal vehicle track in the previous step again under the DTW algorithm, and dividing and judging the normal vehicle track and the outlier track.

Specifically, the normal vehicle track in the previous step is extracted, the similarity between every two tracks is determined based on a DTW algorithm, a similarity matrix S [ a ] [ b ] is listed, the vehicle tracks are divided into two types including normal tracks and outlier tracks through hierarchical clustering, the outlier tracks are determined based on the track number or the dispersion degree epsilon, and therefore the normal vehicle track is extracted.

In general, if the number of the two vehicle trajectories a and B is smaller than that of the other vehicle trajectory, and the ratio of the number to the total number is smaller than an abnormal threshold (generally 30%), the one vehicle trajectory is defaulted as an abnormal outlier trajectory, and the other vehicle trajectory is a normal trajectory; otherwise, the discrete degrees epsilon of the two groups of data are respectively solved, and the group of data with large values is defaulted as the abnormal outlier track.

And 2-2, performing track smoothing on the vehicle track screened and extracted in the step 2-1, re-extracting vehicle track characteristic points, and determining the running range of the vehicle in the single direction.

Step 2-2-1, screening and extracting the normal vehicle track TR in the previous step_jLower trace point p_iRealizing vehicle track smoothing, analyzing and determining new feature point p 'after vehicle track smoothing'_i。

Specifically, a method such as a moving average, a weighted moving average, and an exponential moving average may be used. The invention selects a sliding average smoothing method, which comprises the following steps:

in the formula p_s(i) Representing the smoothed vehicle track points: omega is a smoothing index and is determined according to the number of track points of the shortest track in all tracks and the number of feature points needing to be selected, and the omega takes the value of 5 because 10 feature points are selected again.

Further vehicle trajectory bisectionAnd re-selecting a new characteristic point, and recording the new characteristic point as a vehicle track TR ═ P '| P'_iI is more than or equal to 1 and less than or equal to n, n is the number of the track feature points }, and generally, ten feature points are selected again.

Step 2-2-2, based on all vehicle trajectories TR_jAnd characteristic point p'_iA master curve of vehicle operation is determined. Specifically, feature points p 'are selected based on each vehicle track'_iCorresponding the main curves one by one, determining the central point of the stage i in the same stage i to obtain the main curve TR^midNamely:

in the formula, TR^midRepresenting a vehicle trajectory master curve; p_i^midRepresenting the central point of the i-th stage track feature point, wherein i is more than or equal to 1 and less than or equal to n, and n represents the number of the selected feature points (generally, the value of n is 10); p'_jiAnd representing the new characteristic points of the ith stage track of the jth vehicle track, wherein j is more than or equal to 1 and less than or equal to N, and N represents the total number of the vehicle tracks.

of formula (II b), dist (p'_ji,P_i^mid) And representing the distance between the characteristic point of j tracks i and the central point of the i stage.

And further analyzing the distance numerical values of all track points in the i stage, and defaulting the numerical values which are larger than the normal distribution (mu +2 sigma) to the track stage points as outliers through normal distribution analysis.

Step 2-2-3, according to the main curve TR of the previous step^midAnd on the main curvePhase point P_i^midDetermining envelope point Q of each stage_iAnd then the envelope curve is obtained smoothly.

Step 2-2-3-1, according to stage point P on main curve_i^midDetermining vehicle track points p 'in the same stage'_jiWith main curve TR^midIf the point of the track closest to the main curve is not on the main curve TR^midAnd if so, indicating that the vehicle track points at the stage deviate, and rejecting the vehicle track points, otherwise, turning to the next step.

Step 2-2-3-2, track points p 'on each vehicle track which is not deviated in the same stage'_jiWith main curve TR^midDistance dist (p'_ji,TR^mid) Sorting, selecting 80-90% median value as radius distance of envelope curve corresponding to the main curve at the stage, and recording as r_i。

Step 2-2-3-3, based on the stage point P on the main curve_i^midDetermining envelope point coordinates Q at the same stage_iIs recorded as (x'_i,y′_i). Specifically, the method comprises the following steps:

y′_i＝k_ix′_i+b_i

And

denotes the main curve coordinate in phase i, denoted P_i^mid；k_iRepresents the slope of the envelope in phase i, b_iRepresenting the intercept of the intersection of the envelope with the main curve at stage i.

And 2-2-3-4, judging the position relation between the envelope coordinate and the main curve according to the envelope coordinate solved in the previous step. In particular, the main curve point P according to this phase_i^midAnd the main curve point of the next stage

An envelope threshold ξ is obtained, namely:

and

representing the main curve coordinates on stage i.

X 'if ξ < 0'_iAnd y'_iThe envelope point shown is to the left of the main curve, and ξ > 0 represents x_iAnd y_iThe envelope point is to the right of the main curve, and vice versa on the line where the envelope intersects the main curve.

Step 2-2-3-5, coordinate Q of envelope curve point of all stage points_iIntegrating, smoothing envelope track to obtain new envelope characteristic point q_s(i) To thereby determine the envelope trajectory TR^qNamely, the reasonable driving range of the vehicle track in the mode (direction) is obtained.

And 2-3, repeating the steps 2-1 and 2-2 to determine the track (the reasonable driving range of the vehicle) of each flow direction envelope line of the intersection, and integrating to obtain the reasonable driving range of the vehicle track of the intersection.

Specifically, taking a conventional intersection as an example, the vehicle reasonable driving ranges of the vehicle trajectories of the vehicle turning left, going straight and turning right in four directions are respectively determined.

And 3-1, analyzing and determining the running direction and track information of the vehicle according to the real-time monitoring video.

Step 3-2, comparing the vehicle track path obtained in the step 3-1 with the reasonable driving range of the vehicle track at the intersection obtained in the step 2; and if the vehicle exceeds the driving range, early warning is carried out, and vehicle information is issued to an off-site law enforcement system.

And 3-2, if the ratio of the number of early warning vehicles and the total number of passing vehicles in the running flow direction in unit time is greater than an abnormal condition threshold value (generally about 50%, the value can be selected according to the actual condition of the intersection), early warning is carried out on the flow direction of the intersection, and meanwhile the flow direction of the intersection can be related to a traffic situation monitoring system or a video monitoring system, so that the alarm of the abnormal condition of the intersection is realized, and the abnormal condition of the intersection is monitored.

The following is illustrated by specific examples:

selecting a certain intersection for research, extracting a monitoring video of the intersection within 5min, determining a vehicle track according to track points, and dividing a vehicle track mode by hierarchical clustering, which is specifically shown in fig. 2.

Further extracting and analyzing the vehicle track at the single intersection, taking the left turn of the south entrance as an example, extracting the abnormal track of the vehicle, and determining a main curve and an envelope curve according to the normal vehicle track and the characteristic points thereof, as shown in fig. 3. Namely, the section between the envelope line 1 and the envelope line 2 in the figure is a reasonable driving range of the south-entry left-turn.

When the left-turn vehicle at the south entrance exceeds the reasonable driving range in the process of driving, early warning is carried out, if the early warning frequency in unit time period (1 hour) exceeds 50% of the vehicle-passing record, association is automatically prompted, and the intersection remote monitoring is realized.

The above description is only a preferred embodiment of the present invention, and the scope of the present invention is not limited to the above embodiment, but equivalent modifications or changes made by those skilled in the art according to the present disclosure should be included in the scope of the present invention as set forth in the appended claims.

Claims

1. An abnormal driving early warning method based on a reasonable driving range of vehicles at an intersection is characterized in that: the method comprises the following steps:

step 1, determining each vehicle track of an intersection based on an intersection monitoring historical video, and classifying vehicle track type modes through a hierarchical clustering algorithm and intersection channelized information;

step 2, extracting and analyzing reasonable driving ranges of vehicle tracks in each flow direction of the intersection respectively to realize reasonable range management of vehicles at the intersection; the step 2 specifically comprises:

step 2-1, extracting a vehicle track of a single mode type, and extracting and analyzing a normal vehicle track under the mode type through hierarchical clustering;

step 2-2, performing track smoothing on the basis of the normal vehicle track screened and extracted in the step 2-1, re-extracting vehicle track characteristic points, and determining a reasonable driving range of the vehicle track in a single direction;

step 2-3, repeating steps 2-1 and 2-2 to determine the envelope track of each flow direction of the intersection, and integrating to obtain the reasonable driving range of the vehicle track at the intersection;

the step 2-1 specifically comprises:

step 2-1-1: based on vehicle trajectory TR_jCharacteristic point p_iTrajectory similarity λ to a single vehicle trajectory_jVariance of acceleration

Arc length ratio sigma_jSolving as characteristic data; wherein i is more than or equal to 1 and less than or equal to n, and n is the number of the track characteristic points; j is more than or equal to 1 and less than or equal to N, and N is the number of vehicle tracks;

step 2-1-2: track TR of vehicle_jTrack similarity λ of_jVariance of acceleration

Arc length ratio sigma_jPerforming hierarchical clustering on vehicle tracks serving as feature data, dividing the vehicle tracks into three types, namely abnormal driving tracks, abnormal behavior tracks and normal vehicle tracks, further rejecting the abnormal driving tracks and the abnormal behavior tracks, and realizing normal vehicle track data extraction;

step 2-1-3: performing hierarchical clustering again on the normal vehicle tracks obtained in the step 2-1-2, judging and dividing normal vehicle tracks and outlier tracks based on the track number and the discrete degree of each cluster after clustering, and extracting the normal vehicle tracks;

2. The abnormal driving early warning method based on the reasonable driving range of the vehicles at the intersection as claimed in claim 1, wherein: in the step 1, the concrete steps are as follows:

step 1-1, extracting original track points under different vehicle IDs based on historical videos of intersection monitoring unit time periods, marking as P (f, x, y), wherein f represents a frame number, and x and y represent coordinate values, and further determining single vehicle tracks of different vehicle IDs through initial data cleaning;

step 1-2, classifying vehicle track type modes based on a hierarchical clustering algorithm and intersection channelized information; specifically, the feature point p is performed on a single vehicle track_iAnd extracting, namely performing vehicle track hierarchical clustering through the Euclidean distance or the LCSS distance, eliminating abnormal vehicle tracks, and realizing vehicle track division according to intersection canalization characteristics.

3. The abnormal driving early warning method based on the reasonable driving range of the vehicles at the intersection as claimed in claim 1, wherein:

the step 2-1-1 specifically comprises the following steps: solving the vehicle track TR through an LCSS algorithm or a DTW algorithm according to the vehicle track type_jTrack similarity λ of_j(ii) a Based on vehicle trajectory TR_jCharacteristic point p_iDetermines the vehicle track TR by the number of frames and the distance between the feature points_jVariance of acceleration

Based on vehicle trajectory TR_jDetermining the vehicle trajectory TR by the distance between the characteristic points_jArc length ratio sigma_j；

The step 2-1-2 is specifically as follows: integrating vehicle trajectories TR_jTrack similarity λ of_jVariance of acceleration

Arc length ratio sigma_jThe numerical value is used as a characteristic value, three groups of data are divided by a hierarchical clustering algorithm, and an abnormal driving track, an abnormal behavior track and a normal vehicle track are determined according to the data volume and the dispersion degree epsilon of each group of data; the data set with the minimum data volume and the ratio of the number to the total number smaller than the abnormal threshold value is defaulted as an abnormal driving track, and the data set with the larger dispersion degree epsilon is defaulted as an abnormal behavior track;

and 2-1-3, specifically, performing hierarchical clustering again on the normal vehicle track in the step 2-1-2 based on a DTW algorithm.

4. The abnormal driving early warning method based on the reasonable driving range of the vehicles at the intersection as claimed in claim 1, wherein: the step 2-2 specifically comprises the following sub-steps:

step 2-2-1, normal vehicle track TR1 extracted by screening_mCharacteristic point p of_iRealizing the smoothness of the vehicle track, and equally selecting and determining new characteristic points of the normal vehicle track according to the smoothed vehicle characteristic points;

step 2-2-2, based on all normal vehicle trajectories TR1_mDetermining a main curve of the vehicle operation by the new characteristic points; specifically, new characteristic points selected based on each normal vehicle track are in one-to-one correspondence, and the central point of the stage k is determined under the same stage k to obtain a main curve TR of the vehicle track^midNamely:

in the formula, TR^midRepresenting a vehicle trajectory master curve;

representing the central point of the new characteristic point of the K-th stage track, wherein K is more than or equal to 1 and less than or equal to K, and K represents the number of the selected new characteristic points; p'_mkRepresenting new characteristic points of the kth stage track of the mth normal vehicle track, wherein M is more than or equal to 1 and less than or equal to M, and M represents the number of the normal vehicle tracks;

in the formula (I), the compound is shown in the specification,

representing the distance between the new characteristic point of the kth stage track of the mth normal vehicle track and the central point of the new characteristic point of the kth stage track;

analyzing the distance values of all new feature points of the track in the kth stage, and defaulting the track stage points corresponding to the values larger than the normal distribution as outliers through normal distribution analysis;

step 2-2-3, according to the main curve TR of the vehicle track^midAnd the central point of the new characteristic point of the k-th stage track

Determining envelope point coordinates Q of each stage_kAnd then two envelope lines are obtained smoothly, and the reasonable driving range of the vehicle track is determined.

5. The abnormal driving early warning method based on the reasonable driving range of the intersection vehicle as claimed in claim 4, wherein: the steps 2-2-3 specifically comprise the following sub-steps:

step 2-2-3-1, according to the central point of new characteristic point of k stage track

Determining k stage track new feature point p 'of m normal vehicle track'_mkWith main curve TR of vehicle track^midIf the new characteristic point of the track is not on the main curve TR of the vehicle track at the point where the main curve of the vehicle track is closest to the new characteristic point of the track^midIf the vehicle track is deviated, the vehicle track is rejected, otherwise, the next step is carried out;

step 2-2-3-2, new characteristic points of the track on each vehicle track which is not deviated in the same stage and the vehicle track main curve TR^midDistance dist (p'_mk,TR^mid) Sorting, selecting 80-90% median numerical value as the radius distance of the corresponding envelope curve of the main curve of the vehicle track at the stage, and recording as r_k；

Step 2-2-3-3, based on the central point of the new characteristic point of the k-th stage track

Determining envelope point coordinates Q at the same stage_kNote (x'_k,y'_k) To achieve this, in particular,

y′_k＝K_kx′_k+b_k

in formula (II), x'_kAnd y'_kThe point coordinates representing the envelope at the k-th stage, noted,

and

representing the center point of the new feature point of the k-th stage track

The coordinates of (a); k_kRepresents the slope of the envelope at the k-th stage, b_kRepresenting the intercept of the intersection line of the envelope curve and the vehicle track main curve at the k stage;

step 2-2-3-4, judging the position relation between the envelope coordinate and the vehicle track main curve according to the envelope coordinate solved in the step 2-2-3-3; specifically, the center point of the new feature point is determined according to the k-th stage track

And the central point of the new characteristic point of the k +1 stage track

An envelope threshold ξ is obtained, namely:

xi is a<0 then x'_kAnd y'_kThe point of the envelope curve is shown on the left side of the main curve of the vehicle track, xi>0 represents x'_kAnd y'_kThe envelope curve point is on the right side of the vehicle track main curve, otherwise, the envelope curve point is on a straight line intersecting the vehicle track main curve;

step 2-2-3-5, coordinate Q of envelope point of all stages_kIntegration, versus envelope trajectorySmoothing is carried out to obtain new characteristic points of the envelope line, thereby determining the track TR of the envelope line^qAnd the range formed by the two envelope lines is the reasonable driving range of the vehicle track.

6. The abnormal driving early warning method based on the reasonable driving range of the vehicles at the intersection as claimed in claim 1, wherein: the step 3 comprises the following sub-steps:

step 3-1, determining the running direction and track information of the vehicle according to the real-time monitoring video analysis of the intersection;

step 3-2, comparing the vehicle running direction and track information obtained in the step 3-1 with the reasonable running range of the vehicle track in each flow direction of the intersection obtained in the step 2; if the vehicle exceeds the reasonable driving range of the vehicle track, early warning is carried out, and vehicle information is issued to an off-site law enforcement system;