CN101206229B - A Vehicle Speed Matching Method Based on Linear CCD Image - Google Patents

Abstract

The invention discloses a vehicle speed matching method based on a linear array CCD image. In the method, two linear array CCD cameras are respectively installed on two sections of the road at a certain distance apart. The two linear array CCD cameras point vertically downward, and the linear array CCD cameras The array CCD camera is equipped with an infrared laser source, the frame rate of the line array CCD camera is greater than 1000 frames per second, and the resolution is greater than 1024 lines; the correlation matching calculation is performed on the image sequences obtained by the two line array CCD cameras to obtain the same target The time difference in the two image sequences, using the time difference and the distance between the two cameras, the traveling speed of the target vehicle can be obtained according to the speed formula V=s/t. It is suitable for occasions such as vehicle speed measurement in traffic checkpoints.

Description

A kind of method for matching vehicle speed based on linear array CCD image

Technical field

The invention belongs to traffic detection and technical field of video processing, the method for detecting vehicle speed that relates in the Video Detection is particularly related to a kind of method for matching vehicle speed based on linear array CCD image.

Background technology

For the detection of the travel speed of vehicle, be an important parameter in the traffic detection range, particularly for speed zone, the measurement of the travel speed of vehicle becomes by for important.So the measuring accuracy of the travel speed of raising vehicle is significant.

At present, vehicle speed measurement method relatively more commonly used mainly contains: radar Doppler velocimetry, traffic information collecting instrument velocimetry, video frequency speed-measuring method.Wherein, the radar Doppler velocimetry adopts the principle (Doppler effect: Austrian scientist Doppler discovery of Doppler effect, work as sound, vibration source such as light and radiowave and observer are during with relative velocity v relative motion, the frequency that vibration frequency that the observer received and vibration source are sent is different), when the pulsating wave of the fixing frequency of radar Doppler emission one ran into moving target, the frequency of echo was poor with the emission wave frequency frequency of occurrences, is referred to as Doppler frequency.According to the size of Doppler frequency, can measure the radially speed of related movement of target to radar; According to the mistiming of transponder pulse and acceptance, can measure the distance of target, thus the speed of obtaining.But because the signal emission angle is big, reaching behind the certain distance scattering to be a sector region in a big way, therefore is subjected to the interference of adjacent lane vehicle in use easily, causes the deviation of velocity survey.The traffic information collecting instrument velocimetry passes through along the car body traffic direction, keep at a certain distance away, with the identical measuring system of two-way (2 identical sensors) to identical object, though it is irregular at different time acquired signal sensor output signal, but this two paths of signals is identical, and exist certain hour poor, therefore can ask Vehicle Speed according to the distance and the mistiming of 2 sensors.But its installation need destroy the road surface, reduces the load-bearing capacity on road surface, and inconvenience is installed simultaneously, and maintenance workload is big.Video frequency speed-measuring method with respect to the advantage of above-mentioned several speed-measuring methods is: need not to use specialized equipment, reduce the velocity-measuring system cost greatly, system stability is higher relatively, but traditional video frequency speed-measuring generally adopts area array CCD to carry out, owing to influenced by the not high factor of area array CCD resolution, image frame rate, the rate accuracy of area array CCD is not high.

Summary of the invention

At defective or deficiency that above-mentioned prior art exists, purpose of the present invention exists, and proposes a kind of method for matching vehicle speed based on linear array CCD image, and this method can improve the precision of measuring speed.

In order to realize above-mentioned task, the present invention takes following technical solution:

A kind of method for matching vehicle speed based on linear array CCD image, it is characterized in that, this method is installed two linear array CCD cameras at road respectively on two sections of certain distance, two linear array CCD camera directed in orthogonal are downward, is furnished with the infrared laser source on the linear array CCD camera, the frame rate of linear array CCD camera is greater than 1000 frame/seconds, and resolution is greater than 1024 lines; The image sequence that two linear array CCD cameras are obtained carries out relevant matches calculating, thereby obtain the mistiming of same target in two image sequences, utilize the distance of this mistiming and two video cameras, promptly can obtain the gait of march of target vehicle according to speed formula V=s/t.

The present invention has been owing to adopted ccd video camera, its resolution height, and linear array images sequence vehicle target is clear, have any background interference hardly.Show that through the field experiment application vehicle speed measuring method based on the linear array images sequence of the present invention is effective, the rate accuracy height is applicable to the occasions such as vehicle speed measuring in the traffic block port.

Description of drawings

Fig. 1 is the linear array CCD camera layout;

Fig. 2 is a figure ccd image sequences match target area synoptic diagram;

Fig. 3 is the image sequence of two CCD;

The zone of two image sequences when Fig. 4 is coupling;

Fig. 5 is area segments when being projected as one dimension oscillogram images match;

Fig. 6 is the matching algorithm process flow diagram.

Be described in further detail of the present invention below in conjunction with accompanying drawing.

Embodiment

Method for matching vehicle speed based on linear array CCD image of the present invention, its concrete steps are as follows:

1) on two sections of certain distance (approximately about 2000mm) road, two linear array CCD cameras is being installed, according to linear array CCD video camera imaging characteristics (becoming vertical relation and the composition face by between line and the object that moves), linear array CCD camera direction directed in orthogonal is downward.Select common linear array CCD camera for use, the frame rate of linear array CCD camera is greater than 1000 frame/seconds, and resolution is greater than 1024 lines.

Accompanying drawing 1 is the layout of linear array CCD camera.According to shown in Figure 1, two parallel to each other being installed on the support of certain distance (2 meters) of linear array CCD camera, directed in orthogonal is downward, is furnished with the infrared laser source on the linear array CCD camera, and is the same with the linear array CCD camera direction respectively, vertically downward.The emissive power in infrared laser source is less than 1w.

2) image sequence to two ccd video cameras carries out relevant matches calculating, thereby obtain the mistiming of same target in two image sequences, utilize the distance of this mistiming and two cameras, can obtain the gait of march of target vehicle according to speed formula V=s/t;

Accompanying drawing 2 is to need the zone of mating in two linear array CCD image sequences.Requirement according to coupling, earlier by the CCD1 image sequence is carried out binary conversion treatment, carrying out target then extracts, thereby obtain the target area (seeing the image target area of the line array CCD 1 in the accompanying drawing 2) of CCD1 image sequence, then according to upper limit 180KM/H and the lower limit 36KM/H of the target area in the CCD1 image sequence via speed, obtain when target with speed be 180KM pass through two meters apart from the time, 40ms consuming time.This moment, the target area in the CCD2 image sequence appeared at back 40 row of the target area tail in the CCD1 image sequence, again according to lowest speed 36KM/H, infer the possible position that target occurs in the CCD2 image sequence, be referred to as CCD2 target area (seeing linear array CCD2 image target area in the accompanying drawing 2).The target area M1*N1 of CCD1 image sequence is slided at CCD2 image sequence target area M2*N2, find the moment i (seeing accompanying drawing) of similarity maximum.The position of this moment is the physical location of target in the CCD2 image sequence.

Thereby obtaining the line number of being separated by of two realistic objectives, also is the time.Thereby calculate current goal speed according to speed formula.

Accompanying drawing 3 is depicted as two ccd image sequences.Can find out significantly that from figure in two ccd image sequences, same target vehicle is separated by a distance.

A square frame of drawing with fine rule (what use in the CCD1 image sequence is blue, and what use in the CCD2 image sequence is red) has all appearred in two image sequences in the accompanying drawing 4.Being illustrated in the actual area that finds in the CCD2 image sequence is the zone of getting up with red fine rule frame.

The concrete steps that relevant matches is calculated are as follows:

(a) matching process is to slide in another image sequence according to the target in the image sequence, finds the position in the moment of their similarity maximums, becomes coupling when finding the similarity maximum constantly.Principle according to coupling, the image sequence that first ccd video camera (being called CCD1) collection is come carries out binary conversion treatment earlier, be partitioned into a target area, position according to the coordinate place image sequence of target area, velocity characteristic according to highway car running car, getting 36KM/H is the speed minimum value, and 180KM/H is the maximal value of speed, obtains the probable target area territory of target in the image sequence that second ccd video camera (being called CCD2) collection is come.Two image sequence respective regions are carried out relevant matches to be calculated.

(b) the gray scale function of establishing the CCD1 image sequence is F, and the gray scale function of CCD2 image sequence is G.And the target area size of establishing the CCD1 image sequence is M ₁* N ₁, territory, CCD2 image probable target area size is M ₂* N ₂,, get M according to the territory, probable target area that the velocity limit value is got ₁=M ₂, establish M ₁* N ₁It is capable to start from m, M ₂* N ₂It is capable to start from n, owing to fastest the time, must spend 40ms with the speed of 180KM/H by 2 meters distance, synchronous target appears at the distances that two image sequences differ 40 row at least in the time of also promptly the fastest again. get

n＝m+40 (1)

Have:

F＝f(x，y)x＝1，2，3，…，M1 y＝1，2，3，…，N1 (2)

G＝g(x，y)x＝1，2，3，…，M2 y＝1，2，3，…，N2 (3)

Definition:

$\mathrm{\ρ}(x,y)=\frac{\mathrm{cov}(F,G)}{\sqrt{D_F\·D_G}}---\left(4\right)$

(x y) is the related function of F and G, D to ρ _F, D _GBe respectively the variance of F and G, cov (F G) is the covariance of F and G, has:

$D_F=\frac{1}{\mathrm{MN}}\underset{x=1}{\overset{M}{\mathrm{\Σ}}}\underset{y=1}{\overset{N}{\mathrm{\Σ}}}{(f(x,y)-\stackrel{\‾}{F})}^2,---\left(5\right)$

$D_G=\frac{1}{\mathrm{MN}}\underset{x=1}{\overset{M}{\mathrm{\Σ}}}\underset{y=1}{\overset{N}{\mathrm{\Σ}}}{(g(x,y)-\stackrel{\‾}{G})}^2,---\left(6\right)$

$\mathrm{COV}(F,G)=\frac{1}{\mathrm{MN}}\underset{x=1}{\overset{M}{\mathrm{\Σ}}}\underset{y=1}{\overset{N}{\mathrm{\Σ}}}(f(x,y)-\stackrel{\‾}{F})(g(x,y)-\stackrel{\‾}{\text{G}})\text{---}\left(7\right)$

Be average gray.

Be average gray.

(c) in order to save computing time, under the prerequisite that does not influence formula character, formula is simplified processing, the actual conditions of speed calculation according to the present invention, the key component of taking out formula:

$(f(x,y)-\stackrel{\‾}{F})(g(x,y)-\stackrel{\‾}{G}),---\left(8\right)$

Respectively the target area in two image sequences is carried out the horizontal projection of gray scale, two target areas project into one dimension by bidimensional.Be equivalent to following formula by (8) formula:

$(f\left(u\right)-\stackrel{\‾}{f})(g\left(u\right)-\stackrel{\‾}{g})---\left(9\right)$

F (u) is target area M ₁* N ₁The one dimension function that projects into, length are N ₁,

For at N ₁Gray-scale value is average on the length.According to actual conditions of the present invention, be length after the projection N ₁The zone be N in length ₂The zone in slide, can get maximum correlation.Can obtain effect by (4), (7), (9) formula and follow the same formula of reduction of (7) formula:

$R(F,G)=\underset{u=0}{\overset{N_1}{\mathrm{\Σ}}}(f\left(u\right)-\stackrel{\‾}{f})(g(u+v)-\stackrel{\‾}{g_v})---\left(10\right)$

(d) in order to save the time of coupling, regulation is worked as

In the time of vehicle commander≤200, get

When The vehicle commander gets N greater than 200 the time ₂=200.

(10) in the formula, R (F G) is the sum of products, v=0, and 1 ..., N ₂

Be regional N ₁Gray Projection and mean value,

For sliding into v N constantly ₂Middle length is N ₁Gray Projection and mean value.

Constantly slide, finally obtain R ₁(F, G), R ₂(F, G) ...,

(e) find R _i(F, G) (1≤i≤N ₁) maximal value, this moment i represent the coupling at M ₂* N ₂I in the zone is capable.According to (1) formula, the line number of process is (i+40) line when mating as can be known.

(f) according to the camera sweep frequency, can try to achieve the scanning time that one line spent is (i+40) individual unit.By the distance of time, can try to achieve the instantaneous velocity of target vehicle according to speed formula V=s/t with two cameras.S=200mm, t=(i+40) ms, thus try to achieve the instantaneous velocity of target vehicle.

The applicant carries out horizontal projection to every row in CCD1 image sequence M1*N1 zone, the gray-scale value of each unit in every row is added up, thereby every row adds up and obtains a gray scale accumulated value.After handling like this, the two-dimensional image sequence among the CCD1 projects into one dimension, shown in the accompanying drawing 5 in, blue waveform is that the process of CCD1 image sequence adds up and is projected as oscillogram behind the one dimension.In like manner, projection is carried out in the M2*N2 zone in the CCD2 image sequence, two-dimensional projection becomes one dimension, obtains crocus oscillogram as shown in Figure 5.Article two, between the green line horizontal projection oscillogram of the actual area after the coupling in the CCD2 image sequence.From accompanying drawing 5 as can be seen, it is almost completely identical with blue oscillogram.The proof matching effect is quite desirable, and the actual effect of algorithm is good.

Claims (3)

1. A vehicle speed matching method based on a linear array CCD image, the method installs two linear array CCD cameras respectively on two sections of the road at a certain distance apart, the two linear array CCD cameras point vertically downward, and the linear array CCD The camera is equipped with an infrared laser source, the frame rate of the line array CCD camera is greater than 1000 frames per second, and the resolution is greater than 1024 lines; the correlation matching calculation is performed on the image sequences obtained by the two line array CCD cameras, so as to obtain the same target in two The time difference in the image sequence, utilize this time difference and the distance of two cameras, according to speed formula V=s/t can obtain the traveling speed of target vehicle; It is characterized in that, the specific steps of described correlation matching calculation are as follows: (a) According to the principle of matching, the image sequence collected by the first line array CCD camera is binarized first, and a target area is segmented. Velocity characteristics, take 36KM/H as the minimum value of velocity, and 180KM/H as the maximum value of velocity, obtain the possible target area of the target in the image sequence collected by the second line array CCD camera; carry out the corresponding regions of the two image sequences Correlation matching calculation; (b) Let the grayscale function of the first line array CCD camera image sequence be F, and the grayscale function of the second line array CCD camera image sequence be G; and set the target area of the first line array CCD camera image sequence The size is M ₁ ×N ₁ , the size of the possible target area of the second line array CCD camera image is M ₂ ×N ₂ , the possible target area obtained according to the speed limit value, take M ₁ =M ₂ , let M ₁ ×N ₁ It starts at line m, M ₂ ×N ₂ starts at line n, and because of the fastest speed, the time required to pass a distance of 2 meters at a speed of 180KM/H is 40ms, that is, the synchronization target appears at the fastest time The distance between the two image sequences is at least 40 lines; n=m+40; have: F=f(x,y)x=1, 2, 3,..., M1y=1, 2, 3,..., N1 (2) G=g(x,y)x=1, 2, 3,..., M2y=1, 2, 3,..., N2 (3) definition:

ρ(x, y) is the correlation function between F and G, D _F , D _G are the variances of F and G respectively, cov(F, G) is the covariance of F and G, there are:

is the average gray value;

is the average gray value; (c) Simplify the formula, and take out the key part of the formula according to the actual situation of speed calculation:

The horizontal projection of the gray level is performed on the target areas in the two image sequences respectively, and the two target areas are projected from two dimensions to one dimension, and the formula (8) is equivalent to f(u) is a one-dimensional function projected into the target area M ₁ ×N ₁ , the length is N ₁ ,

is the average of the gray value on the length of N ₁ ; according to the actual situation, slide the area of length N ₁ after projection in the area of length N ₂ , and the relevant maximum value can be obtained; From formulas (4), (7) and (9), a simplified formula with the same effect as formula (7) can be obtained

(d) In order to save matching time, specify when

when, take

when

When it is greater than 200, take N ₂ =200; In formula (10), R(F, G) is the sum of products, v=0, 1, ..., N ₂

is the average value of the area N ₁ grayscale projection sum,

is the average value of the gray-scale projection sum of length N ₁ in N ₂ at time v; continuous sliding, and finally get R ₁ (F, G), R ₂ (F, G), ...,

(e) Find the maximum value of R _i (F, G) (1≤i≤N ₁ ), at this time, i means matching the i-th row in the M ₂ ×N ₂ area. According to formula (1), it can be seen that the matching The number of lines passed by is (i+40) lines; (f) According to the scanning frequency of the linear array CCD camera, the time required to scan a line can be obtained as (i+40) units, and the distance between the time and the two linear array CCD cameras, according to the speed formula V=s/t, The instantaneous speed of the target vehicle can be obtained. 2. The method according to claim 1, wherein the direction of the infrared laser source is the same as the direction of the linear array CCD camera, vertically downward. 3. The method according to claim 1, wherein the emission power of the infrared laser source is less than 1w. the

Images