CN104794425A - Vehicle counting method based on movement track - Google Patents

Abstract

The invention discloses a statistical method based on vehicle tracks, which relates to the field of traffic detection. The invention includes the selection and tracking of stable feature points, clustering is performed by using the relationship changes between the relative distances of the stable feature points in the real space trajectory at the same moment, and when the variance of the relative distance at the same moment is less than the preset threshold, no update The count value of the vehicle counter, when the variance of the relative distance at the same time is greater than a preset threshold, the count value of the vehicle counter is increased by one. The present invention extracts the vehicle trajectory in the captured road video image, processes the vehicle trajectory therein, obtains the relative distance of the trajectory at the same time, and determines the number of vehicles in the video image according to the variation relationship of the relative distance at the same time, so that In the process of counting the number of vehicles, the influence of the environment and the limitation that only a single vehicle type can be counted are avoided, and the statistical precision and efficiency of the number of vehicles are improved.

Description

A Vehicle Statistical Method Based on Trajectory

technical field

The invention relates to the field of traffic detection, in particular to a statistical method based on vehicle tracks.

Background technique

With the rapid development of the economy, the number of cars is increasing rapidly. In order to reduce the traffic congestion caused by the increase in the number of cars, it is necessary to obtain the real-time number of cars on the road for timely traffic management.

In the existing technology, the methods for vehicle statistics mainly include the ground induction coil method, wave frequency detection method, etc., wherein the ground induction coil method is that the driving vehicle passes through the road surface buried with coils, causing the change of the coil magnetic field, so that the detector Obtain the number of the vehicle; and the wave frequency detection method is to launch continuous low-power modulated electromagnetic waves into a certain area, leaving a projection on the road surface, and divide the projection area into multiple layers according to the specified distance. When the vehicle enters the After the projected area, electromagnetic waves will be reflected and then received by the receiver. According to the car manufacturer of the fixed model and the time when the reflected electromagnetic wave is received, the number of vehicles passing through the projected area is estimated.

In the process of realizing the present invention, the inventor finds that there are at least the following problems in the prior art:

The coil using the ground induction coil method is buried underground, which is easily affected by freezing, subsidence of the roadbed, saline-alkali and other environments, which will affect the accuracy; although the accuracy of the wave frequency detection method will not be affected, but due to the frequency of the emitted electromagnetic wave Single, only one type of vehicle type can be counted. Once multiple types of vehicle types appear, accurate statistics cannot be carried out, which reduces the statistical efficiency.

Contents of the invention

In order to solve the problems of the prior art, the present invention provides a vehicle statistics method based on driving trajectory, the method is used to process the captured road video images, and the method includes:

Select a first number of sample points in the video image, establish a first coordinate system in the video image, establish a second coordinate system in real space, and determine the first number of sample points at the first coordinates Coordinates in the first coordinate system and coordinates in the second coordinate system, determine a coordinate transformation matrix according to the coordinates of the first number of sample points in the first coordinate system and the coordinates in the second coordinate system;

Processing the video image to obtain a binarized image;

Select feature points in the video image, determine the coordinates of the feature points in the first coordinate system, determine the foot point corresponding to the feature points in the binarized image, and determine the feature points. the coordinates of the foot point in the first coordinate system;

According to the coordinates of the feature points in the first coordinate system, combined with the coordinate transformation matrix, obtain the coordinates of the feature points in the second coordinate system, and according to the coordinates of the feature points in the second coordinate system Coordinates in the system to determine stable feature points;

Obtaining the coordinates of the stable feature points in the first coordinate system in each frame of the video image, and combining the coordinate transformation matrix to determine the continuous coordinates of the stable feature points in the second coordinate system, determining a movement trajectory according to the continuous coordinates;

Select a reference detection line in the second coordinate system, and when the movement trajectory passes the reference detection line, perform filtering processing on the height values of the stable feature points in the movement trajectory to obtain a processed mean value ;

Obtain multiple sets of relative distances at the same time of the moving track, when the variance of the relative distance at the same time is greater than a preset threshold, update the count value of the vehicle counter, and sequentially compare the relative distance at the same time of the moving track to be acquired Judgment is made until the judgment ends, and the number of vehicles in the video image is determined according to the count value of the vehicle counter.

Optionally, the processing of the video image to obtain a binarized image includes:

Remove the background of each frame of image in the video to obtain an initial binarized image with only the vehicle subject;

Connected domain analysis and hole filling are performed on the initial binarized target image to obtain a processed binarized image.

Optionally, determining the foot point corresponding to the feature point in the binarized image includes:

In the binarized image, starting from the feature point, combined with the first coordinate system, a vertical line is drawn in the vertical direction, and the intersection between the vertical line and the binarized image There is an intersection point at , and let the intersection point be the foot point corresponding to the feature point.

Optionally, the determining stable feature points according to the coordinates of the feature points in the second coordinate system includes:

In the second coordinate system, determine the height value of the feature point;

When the height value is less than a preset height threshold, the feature point is the stable feature point.

Optionally, after the determination of the movement trajectory, the method further includes:

The moving track is smoothed by Kalman filtering, and the moving track is corrected according to the smoothed track to obtain a corrected track.

Optionally, performing filtering processing on the height values of the stable feature points in the moving trajectory to obtain a processed mean value, and correcting the moving trajectory according to the mean value, including:

Extracting the height values of all the stable feature points in the moving track;

Obtain the average value of all said height values;

The moving track is corrected according to the average value.

Optionally, after the acquisition of the relative distances of multiple sets of the moving tracks, the method further includes:

When the variance of the relative distance at the same time is less than a preset threshold, mark the two moving trajectories corresponding to the relative distance as the same group, and do not update the count value of the vehicle counter; otherwise, when the same time When the variance of the relative distance is greater than the preset threshold, mark the two moving trajectories corresponding to the relative distance as different groups, add one to the count value of the vehicle counter, and continue to obtain the relative distance of the moving trajectories. The distance is judged until the end of the judgment.

The beneficial effects brought by the technical scheme provided by the invention are:

By extracting the vehicle trajectory in the captured road video image, and processing the vehicle trajectory, the relative distance of the trajectory at the same time is obtained, and the number of vehicles in the video image is determined according to the change relationship of the relative distance at the same time, so that in the vehicle The process of counting the number avoids the influence of the environment and the limitation that only a single vehicle type can be counted, and improves the statistical accuracy and efficiency of the number of vehicles.

Description of drawings

In order to illustrate the technical solution of the present invention more clearly, the accompanying drawings that need to be used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. Ordinary technicians can also obtain other drawings based on these drawings on the premise of not paying creative work.

Fig. 1 is a schematic flow chart of a vehicle statistics method based on driving trajectory provided by the present invention;

Fig. 2 is the schematic diagram of selecting sample points in the statistical method provided by the present invention;

Fig. 3 is the binary image obtained in the statistical method provided by the present invention;

Fig. 4 is a schematic diagram of determining feature points and corresponding feet in a binarized image in the statistical method provided by the present invention.

Fig. 5 is a schematic diagram of a moving track formed by a plurality of stable feature points in the statistical method provided by the present invention;

Fig. 6 is a schematic diagram of determining detection lines in the statistical method provided by the present invention;

Fig. 7 is a schematic diagram of analyzing the acquired movement trajectory in the statistical method provided by the present invention.

detailed description

In order to make the structure and advantages of the present invention clearer, the structure of the present invention will be further described below in conjunction with the accompanying drawings.

Embodiment one

The present invention provides a vehicle statistics method based on driving trajectory, the method is used to process the captured road video images, as shown in Figure 1, the method includes:

101. Select the first number of sample points in the video image, establish the first coordinate system in the video image, establish the second coordinate system in the actual space, and determine the coordinates of the first number of sample points in the first coordinate system and the coordinates in the second coordinate system, and determine a coordinate transformation matrix according to the coordinates of the first number of sample points in the first coordinate system and the coordinates in the second coordinate system.

In implementation, the first coordinate system is established based on the video image, that is, the X-Y plane coordinate system, and the second coordinate system is established based on the actual scene, that is, the X-Y-Z three-dimensional coordinate system, where the first number of sample points is at least are six points with known coordinates, at least four of which are points on the plane coordinate system, and at least two are points with height values (that is, non-zero coordinate values on the Z axis) in the solid coordinate system.

For the convenience of description, the six point positions selected in this embodiment are shown in Figure 2, where the coordinate values of (i, j, k, l, m, n) in the second coordinate system are i=(15,0 ,0), j=(15,15,0), k=(0,15,0), l=(0,0,0), m=(0,0,1.1), n=(0,15 ,1.1), the coordinate values of (i, j, k, l, m, n) in the first coordinate system are related to the position in the image. The conversion matrix C is determined by the corresponding relationship between (i, j, k, l, m, n) in the first coordinate system and the second coordinate system.

The calculation formula of the above conversion matrix C is:

$\left[\begin{array}{c}\mathrm{<span\; class="notranslate">\; u</span>}\\ \mathrm{<span\; class="notranslate">\; v</span>}\\ \mathrm{<span\; class="notranslate">\; l</span>}\end{array}\right]<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; C</span>}\left[\begin{array}{c}\mathrm{<span\; class="notranslate">\; x</span>}\\ \mathrm{<span\; class="notranslate">\; the\; y</span>}\\ \mathrm{<span\; class="notranslate">\; z</span>}\\ \mathrm{<span\; class="notranslate">\; l</span>}\end{array}\right]<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

The point Q corresponding to the coordinates (x, y, z, 1) is a point in the second coordinate system, and the point q corresponding to the coordinates (u, v, 1) is the point corresponding to Q in the first coordinate system. The above coordinates use homogeneous coordinates because in the process of obtaining the transformation matrix C, conversion between the camera coordinate system, the image coordinate system and the world coordinate system is required. Using homogeneous coordinates facilitates the conversion between coordinate systems.

102. Process the video image to obtain a binarized image.

This step specifically includes:

Remove the background of each frame of image in the video to obtain an initial binarized image with only the vehicle subject;

Connected domain analysis and hole filling are performed on the initial binarized target image to obtain the processed binarized image.

In implementation, for each frame image in the video image, do the following processing:

First, identify the target subject for each frame of the video image.

Optionally, the target subject of the image in this embodiment is a single vehicle.

Secondly, in each frame of image, the background except the target subject is removed, and the remaining images with the target subject are binarized to obtain the initial binarized image.

Optionally, after confirming the target subject in each frame of image, the background part except the target subject can be set to black, and the target subject can be set to white, so that an initial binarized image is obtained. Of course, other binarization algorithms can also be used here to obtain a similar initial binarization image.

Finally, the connected domain analysis and hole filling are performed on the initial binary image to obtain the processed binary image.

Optionally, due to algorithm limitations in the previously obtained initial binary image, features such as burrs and cut-offs will be left in the initial binary image. For the above-mentioned features that exist, this can use connected domain analysis and hole filling The method removes the existing features such as burrs and partitions, and obtains a clearer and clearer binarized image than the original binarized image. The binarized image is shown in Figure 3.

103. Select feature points in the video image, determine the coordinates of the feature points in the first coordinate system, determine the foot points corresponding to the feature points in the binarized image, and determine that the foot points are in the first coordinate system coordinate of.

Wherein, determining the foot point corresponding to the feature point in the binarized image includes:

In the binarized image, starting from the feature point, combined with the first coordinate system, a vertical line is drawn in the vertical direction. There is an intersection point between the vertical line and the binarized image, and the intersection point is the feature point corresponding to the foot point.

In the implementation, the feature point is selected from the video image, and the coordinates of the feature point in the first coordinate system are obtained. In the binarized image as shown in Figure 4, the determined feature point is f, and at the first coordinate In the system, a vertical line is drawn along the vertical direction from the feature point, and the vertical line intersects with the boundary of the target image at point b, which is the vertical foot corresponding to the feature point f.

104. Obtain the coordinates of the feature points in the second coordinate system according to the coordinates of the feature points in the first coordinate system combined with the coordinate transformation matrix, and determine the stable feature points according to the coordinates of the feature points in the second coordinate system.

Among them, according to the coordinates of the feature points in the second coordinate system, the stable feature points are determined, including:

In the second coordinate system, determine the height value of the feature point;

When the height value is less than the preset height threshold, the feature point is a stable feature point.

In the implementation, from the image coordinates (x, y) of the feature points, the transformation matrix C and the plane coordinates (X, Y) of the X-Y axis plane coordinate system of the feature points, the coordinate values of the feature points in the 3D space can be obtained as (X,Y,Z), for the calculation process of the height value Z, as follows:

In Fig. 4, f=(x, y) is the feature point, b=(u, v) is the foot point; according to the known transformation matrix C, it is brought into the formula (2)

$\left[\begin{array}{c}\mathrm{<span\; class="notranslate">\; u</span>}\\ \mathrm{<span\; class="notranslate">\; v</span>}\\ \mathrm{<span\; class="notranslate">\; l</span>}\end{array}\right]<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; C</span>}\left[\begin{array}{c}\mathrm{<span\; class="notranslate">\; x</span>}\\ \mathrm{<span\; class="notranslate">\; Y</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; l</span>}\end{array}\right]<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

Obtain the coordinates (X, Y, 0) of the foot point b corresponding to the second coordinate system; and because the stable feature point and the foot point are on the same vertical line in the second coordinate system, so in the second coordinate system The (X, Y) coordinate values of the feature point F and the foot point B are the same, and the coordinate values (x, y) of the known feature point in the first coordinate system are brought into the formula (3)

$\left[\begin{array}{c}\mathrm{<span\; class="notranslate">\; x</span>}\\ \mathrm{<span\; class="notranslate">\; the\; y</span>}\\ \mathrm{<span\; class="notranslate">\; l</span>}\end{array}\right]<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; C</span>}\left[\begin{array}{c}\mathrm{<span\; class="notranslate">\; x</span>}\\ \mathrm{<span\; class="notranslate">\; Y</span>}\\ \mathrm{<span\; class="notranslate">\; Z</span>}\\ \mathrm{<span\; class="notranslate">\; l</span>}\end{array}\right]<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>$

Get the height value Z.

After the height value Z of the feature point is determined according to the above steps, if the height value Z is less than the preset height threshold Th, the feature point corresponding to the height value Z is used as a stable feature point; if the height value Z is greater than or equal to the preset If the height threshold Th is set, the feature point is discarded, and the comparison of the next feature point is continued until all feature points are compared.

105. Acquire the coordinates of the stable feature points in the first coordinate system in each frame of video image, determine the continuous coordinates of the stable feature points in the second coordinate system in combination with the coordinate transformation matrix, and determine the movement trajectory according to the continuous coordinates.

Further, after determining the movement track, the method also includes:

The moving trajectory is smoothed by Kalman filter, and the moving trajectory is corrected according to the smoothed trajectory to obtain the corrected trajectory.

In implementation, this step 105 is specifically described as follows:

Firstly, the coordinates of the stable feature point in the first coordinate system in each frame of video image are obtained, combined with the transformation matrix C, the coordinates of the stable feature point in the second coordinate system are obtained.

Secondly, according to the above method, the continuous coordinates of the stable feature point in the multi-frame video images are determined, and the moving track of the stable feature point is obtained according to the continuous coordinates, the continuous track is shown in FIG. 5 .

Finally, the Kalman filter is used to smooth the obtained moving trajectory, and the moving trajectory is corrected according to the smoothed trajectory to obtain the corrected trajectory.

Optionally, the Kalman filter here is to evaluate the coordinate value of each point in the moving track, and replace the coordinate values that obviously deviate from the average level according to the average level of multiple coordinate values in the direction of each coordinate axis, As a result, the coordinate values in the direction of each coordinate axis are basically close to unity, so that the trajectory corresponding to the corrected coordinate values will be smoother. Of course, only the Kalman filter is used for trajectory correction here, and other filtering methods can also be used for similar processing, which will not be repeated here.

106. Select a reference detection line in the second coordinate system, and when the moving trajectory passes the reference detection line, filter the height values of the stable feature points in the moving trajectory to obtain the processed average value, and correct the moving trajectory according to the average value .

Among them, the height value of the stable feature point in the moving track is filtered to obtain the processed mean value, including:

Extract the height values of all stable feature points in the moving trajectory;

Get the average of all height values;

The movement trajectory is corrected according to the average value.

In implementation, first, similar to the calculation process in step 104, the coordinates of all stable feature points on the i-th image trajectory in the second coordinate system are determined using the transformation matrix matrix C These coordinates come with errors.

Secondly, a detection line located on the road surface and perpendicular to the lane line is selected as the reference detection line. The reference detection line is shown as the dotted line in Figure 6 in the actual scene. When a moving trajectory passes the detection line, the Median filtering is performed on the height information of all stable feature points to obtain the height of the trajectory as That is, the obtained height z _i is the average value of all height values.

Finally, the coordinate value of each stable feature point on the trajectory in the second coordinate system is corrected by using the above average height z _i , and the corrected coordinates are (x _i1 , y _i1 , z _i )...( x _in , y _in , z _i ), which reduces the error caused when using the foot point to find the coordinates of the feature point in the second coordinate system.

107. Obtain the relative distances of multiple groups of moving trajectories at the same time. When the variance of the relative distances at the same time is greater than the preset threshold, update the count value of the vehicle counter, and sequentially compare the variance of the relative distances at the same time of the acquired moving trajectories. Judging until the end of the judgment, according to the count value of the vehicle counter, determine the number of vehicles in the video image.

Among them, after obtaining the relative distances of multiple sets of moving trajectories, it also includes:

When the variance of the relative distance at the same moment is less than the preset threshold, mark the two moving trajectories corresponding to the relative distance as the same group, and do not update the count value of the vehicle counter; otherwise, when the variance of the relative distance at the same moment is greater than the preset threshold , mark the two moving trajectories corresponding to the relative distance as different groups, add one to the count value of the vehicle counter, and continue to judge the relative distance of the acquired moving trajectories until the judgment ends.

In practice, the relative distance here refers to the relative distance of the two trajectory lines at the same moment. Select multiple consecutive moments to obtain multiple relative distances. The variance of the relative distance is used as the basis for measuring the change of the relative distance. The relative distance When the variance of is less than the preset threshold, it means that the relative distance between the two trajectories at the same moment during the tracking process remains basically unchanged, and they belong to the same vehicle, and the value of the vehicle counter remains unchanged; otherwise, when the relative distance When the variance of is greater than the preset threshold, it means that the relative distance between the two trajectories changes greatly at the same time during the tracking process, and the trajectories do not belong to the same vehicle, and the value of the vehicle counter is increased by one.

The statistical diagram is shown in Figure 7, where the trajectory line with a marker value of 0 represents a vehicle, and the variance of the relative distance d between two trajectory lines with a marker value of 1 at the same time is 1.2, which is less than Th2, and the threshold value Th2 is 2.5 , these two trajectory lines represent a vehicle, and the trajectory line with a marker value of 2 represents another vehicle; count all vehicles passing the detection line in the detection area.

The idea of "clustering" is used to determine whether the two trajectories are the trajectories of the same vehicle according to the relative distance. That is, get the first track, mark 1, get the second track, and calculate the variance of the relative distance at the same time as the first track, if the variance of the relative distance at the same time is less than the threshold, mark the second track as 1, if at the same time The variance of the relative distance is greater than the threshold, marked as 2, and the counter is incremented by 1, that is, every time a new trajectory is obtained, the relative distance variance is judged at the same time as all the previous trajectories, and the operation of adding 1 to the counter is performed according to the result, until Judgment ends, counting ends.

In this step, the reason why we have been emphasizing the comparison of "relative distance at the same moment" is because in the collected video images, each frame of image corresponds to the situation of the road at the same moment, using "simultaneous moment "Description is a description of the real situation; in addition, only by comparing the relative distances of the trajectories at the "simultaneous moment" can the movement of a vehicle be determined, otherwise, even the trajectory of a vehicle will The situation will be different, and the meaning of judging the number of vehicles based on the trajectory will be lost.

A vehicle statistics method based on driving trajectory proposed in this embodiment extracts the vehicle trajectory in the captured road video images and processes the vehicle trajectory to obtain the relative distance of the trajectory at the same time, according to the relative distance at the same time The variation relationship of the distance determines the number of vehicles in the video image, so that the influence of the environment and the limitation of only a single vehicle type can be counted in the process of counting the number of vehicles, and the statistical accuracy and efficiency of the number of vehicles are improved.

It should be noted that the statistical method provided by the above-mentioned embodiment is only used as an illustration of the actual application of the statistical method, and the above-mentioned statistical method can also be used in other application scenarios according to actual needs. The specific implementation process is similar to the above embodiment, and will not be repeated here.

The above description is only an embodiment of the present invention, and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention Inside.

Claims (7)

1., based on a statistical method for track of vehicle, described method is used for processing the road video image photographed, and it is characterized in that, described method comprises:

The sample point of the first quantity is chosen in described video image, the first coordinate system is set up in described video image, the second coordinate system is set up in real space, determine the described coordinate of the first number of samples point in described first coordinate system and the coordinate in described second coordinate system, according to the coordinate of described first number of samples point in described first coordinate system and the coordinate determination coordinate conversion matrix in described second coordinate system;

Described video image is processed, obtains the image after binaryzation;

Selected characteristic point in described video image, determines the coordinate of described unique point in described first coordinate system, determines the intersection point point with described Feature point correspondence, determine the coordinate of described intersection point point in described first coordinate system in the image after described binaryzation;

According to the coordinate of described unique point in described first coordinate system, in conjunction with described coordinate conversion matrix, obtain the coordinate of described unique point in described second coordinate system, according to the coordinate of described unique point in described second coordinate system, determine invariant feature point;

Obtain the coordinate in described first coordinate system of described invariant feature point in video image described in every frame, in conjunction with described coordinate conversion matrix, determine the continuous coordinate of described invariant feature point in described second coordinate system, according to described continuous coordinate, determine motion track;

Benchmaring line is chosen in described second coordinate system, at described motion track through described benchmaring line, filtering process is carried out to the height value of the described invariant feature point in described motion track, obtains the average after processing, according to described average, described motion track is revised;

Obtain the relative distance in the same time of the described motion track of many groups, when the variance of described relative distance is in the same time greater than predetermined threshold value, upgrade the count value of vehicle counter, successively the variance of the relative distance in the same time by the described motion track got is judged, until judge to terminate, according to the described count value of described vehicle counter, determine the number of vehicles in described video image.

2. method according to claim 1, is characterized in that, is describedly processed by described video image, obtains the image after binaryzation and comprises:

By the background removal of each two field picture in video, only be there is the initial binary image of vehicle body;

Connected domain analysis and holes filling are carried out to described initial binary target image, obtains the image after the binaryzation after processing.

3. method according to claim 1, is characterized in that, determines, with the intersection point point of described Feature point correspondence, to comprise in the image after described binaryzation:

In image after described binaryzation, initial from described unique point, in conjunction with described first coordinate system, do vertical line to vertical direction, there is an intersection point in the intersection in the image after described vertical line and described binaryzation, makes described intersection point be the described intersection point point with described Feature point correspondence.

4. method according to claim 1, is characterized in that, described according to the coordinate of described unique point in described second coordinate system, determines invariant feature point, comprising:

In described second coordinate system, determine the height value of described unique point;

When described height value is less than default height threshold, described unique point is made to be described invariant feature point.

5. method according to claim 1, is characterized in that, described determine motion track after, described method also comprises:

Utilize Kalman filtering to the smoothing process of described motion track, according to the track after described smoothing processing, described motion track is revised, obtain revised track.

6. method according to claim 1, is characterized in that, the described height value to the described invariant feature point in described motion track carries out filtering process, obtains the average after processing, revises, comprising according to described average to described motion track:

Extract the height value of whole described invariant feature point in described motion track;

Obtain the mean value of whole described height value;

According to described mean value, described motion track is revised.

7. method according to claim 1, is characterized in that, organize the relative distance in the same time of described motion track in described acquisition after, described method also comprises more:

When the variance of described relative distance is in the same time less than predetermined threshold value, corresponding for described relative distance two described motion tracks is labeled as same group, does not upgrade the count value of described vehicle counter; Otherwise, when the variance of described relative distance is in the same time greater than predetermined threshold value, corresponding for described relative distance two described motion tracks are labeled as different groups, the count value of described vehicle counter is made to add one, continue the relative distance of the described motion track got to judge, until judge to terminate.