CN104575003A - Method for detecting vehicle speed based on road monitoring videos - Google Patents

Abstract

The invention discloses a method for detecting vehicle speed based on road monitoring videos. The method comprises the following steps: (1) dividing a lane using a lane marking line identification method and eliminating unnecessary background; (2) obtaining lane width information according to lane images; (3) calculating the actual distance represented by each pixel point; (4) solving the correction factor Beta which is used for correcting the actual lane distance represented by each pixel point; (5) calculating the optical flow of the vehicle to obtain the trajectory of the vehicle; (6) calculating the actual moving distance of the vehicle; (7) multiplying the correcting the distance by the correction factor to correct the distance; (8) calculating the speed. According to the invention, a video image analysis technology is adopted to decrease the cost of vehicle speed monitoring and speeding behavior recording; and in addition, the method also has the function of expanding the vehicle speed monitoring scope, and vehicle speed monitoring can be implemented at any place with cameras.

Description

A kind of vehicle speed detection method based on traffic surveillance videos

Technical field

The present invention relates to intelligent traffic vehicle to test the speed field, particularly relate to the Vehicle Velocity Measurement Method based on traffic surveillance videos image.

Background technology

Along with China's expanding economy, the popularity rate of automobile also rapidly promotes thereupon, and while vehicle increases, traffic hazard also comes one after another.And to drive over the speed limit be one of arch-criminal of leading to of tragedy.

The main method of current vehicle speed measuring has radar velocity measurement, laser velocimeter, ground sensing coil speed measuring etc.Three kinds of speed-measuring methods all have the feature of high-accuracy, but cannot carry out record to unlawful practice while measuring speed.Especially, ground sensing coil speed measuring needs at embedded underground coil, and road pavement damage is comparatively large and maintenance cost is higher.

Summary of the invention

The present invention is and overcomes the above problems and produce, and being provides a kind of vehicle speed detection method based on traffic surveillance videos, is intended to the loaded down with trivial details degree and the cost that reduce vehicle speed measuring.The present invention is based on the vehicle speed measuring method of traffic video, its external hardware device only depends on video camera, and software and hardware cooperation can solve tests the speed and monitor two pieces of difficult problems.

Innovative point of the present invention is: the number of pixels (pixel_number) proposing lane width in actual range (pixel_distance)=track developed width (the road_width)/video image representated by each pixel.

Country has formulated a set of detailed standard for the width in track, and normalized lane width is for the invention provides great facility.

Because the lane width in video image constantly changes along with the shooting degree of depth of camera, therefore distance representated by pixel is also relevant with the position at pixel place.If pixel coordinate is (x, y), then when y=i(establishes the longitudinal width of each pixel consistent with the distance representated by transverse width) time:

pixel_distance[i]=road_width/pixel_number[i]。

Based on the vehicle speed detection method of traffic surveillance videos, it is characterized in that, comprise the steps:

1. track picture intercepts: adopt track marking line method of identification from video image, be partitioned into a track.Delete superfluous background, obtains pure track picture.Original image element used during sectional drawing is consistent with video pixel.Also Manual interception method can be adopted to obtain.

2. obtain lane width information: the track picture 1. obtained according to step, calculate lane width, represent by pixel number (pixel_number).

3. the actual range representated by every pixel is calculated:

Pixel_distance [i]=road_width/pixel_number [i], stored in document, facilitates subsequent calls by one group of pixel_distance [i] value obtained.

4. solve correction factor β: carry out distance calibration to track, more real distance and the gap calculating gained distance, as correction factor, in order to correct the actual track distance representated by each pixel.Namely actual range/calculating the gained two between β=two mark marks spacing.

5. registration of vehicle movement locus: suppose that vehicle does rectilinear motion, adopts feature point tracking mode to obtain the light stream of vehicle movement, and the average calculating all light streams is used for being similar to the movement locus replacing vehicle.

6. vehicle actual motion distance is calculated: if there is not the change in direction in the driving process of vehicle, then according to vehicle actual motion distance can be calculated stored in the pixel_distance [i] in document and vehicle movement pixel number, this distance values equals the actual motion distance of the actual range sum in vehicle movement track representated by each pixel as vehicle.That is: distance solution formula sum formula replaces integral formula; Formula is $\underset{i=y1}{\overset{y2}{\mathrm{\Σ}}}\mathrm{pixel}\_\mathrm{dis}\tan \mathrm{ce}\left[i\right].$ pixel_distance[i]。

7. distance correction: vehicle actual motion distance is multiplied by correction factor β;

8. speed calculates: according to the move distance of vehicle in reality and the actual motion speed of run duration calculating vehicle.Namely wherein T is front and back two frame video image frame difference/video frame rates.

Further, step 6. in, there is the change in direction in the traveling because of vehicle, therefore the deviation angle that the actual range of gained and vehicle travel should be combined, to obtain more accurate result.If step 5. in the first and last point pixel coordinate of vehicle movement track of record be respectively (x1, y1), (x2, y2), the deviation angle that can obtain in vehicle travel process is $\mathrm{\α}=\sqrt{{(x1-x2)}^2+{(y1-y2)}^2}/|y2-y1|,$ Then the final computing formula of vehicle actual motion distance is:

$D=\underset{i=y1}{\overset{y2}{\mathrm{\Σ}}}\mathrm{pixel}\_\mathrm{dis}\tan \mathrm{ce}\left[i\right]*\mathrm{\β}*\mathrm{\α}.$

The present invention adopts video image analysis technology, and the simple consumptive material of method is single, only needs a camera.Can carry out car speed monitoring after photographed data being passed back Surveillance center, Detection results is excellent.The movement velocity of vehicle can be obtained after adopting above-mentioned steps of the present invention.The present invention can reduce the cost of vehicle speed detection and hypervelocity behavior record.In view of the popularity rate of camera is higher, so the monitoring can carrying out in a big way to vehicle, can report to the police or submit to car pipe office to punish it if there is hypervelocity behavior, play the effect of supervision driving personnel, effectively can reduce the hypervelocity behavior of vehicle, ensure the security of the lives and property of driving personnel and the masses.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of vehicle speed detection step of the present invention.

Embodiment

In order to explain technical scheme of the present invention further, below by specific embodiment, detailed explanation is carried out to the present invention.

1. track picture intercepts: the pure road picture intercepting a track from video image, as shown in Figure 1.

2. obtain lane width information: set the coordinate of each pixel in carriageway image as (x, y), then, as y=i, obtain lane width, represent by pixel number, i.e. pixel_number [i].

3. calculate the track actual range representated by every pixel: if track real wide be 3.5 meters, to be then the actual track distance representated by some row pixels of (x, i) be coordinate points:

pixel_distance[i]=3.5m/pixel_number[i]。By the actual range pixel_distance [i] representated by each pixel stored in document, the vehicle speed measuring after being convenient to.

4. solve correction factor: distance calibration is carried out to track, such as, set a mark every 5m.If step 3. gained data calculate two mark between distance be z rice, then correction factor β=5/z.

5. registration of vehicle movement locus: adopt feature point tracking mode to obtain the light stream of vehicle movement, the average calculating all light streams is used for being similar to the movement locus replacing vehicle.

6. calculate vehicle actual motion distance D: establish the first and last point coordinate of vehicle movement track to be respectively (m1, n1), (m2, n2), the deviation angle obtaining vehicle traveling is:

$\mathrm{\α}=\sqrt{{(x1-x2)}^2+{(y1-y2)}^2}/|y2-y1|,$ Then the final computing formula of vehicle actual motion distance is: $D=\underset{i=y1}{\overset{y2}{\mathrm{\Σ}}}\mathrm{pixel}\_\mathrm{dis}\tan \mathrm{ce}\left[i\right]*\mathrm{\β}*\mathrm{\α}.$

7. distance correction: vehicle actual motion distance is multiplied by correction factor and obtains vehicle movement distance in reality.

8. speed calculates: set the frame between two width video images poor as k, the acquisition frame rate of CCTV camera is F, then the speed of a motor vehicle is: $v=\frac{D}{T}=\frac{D*F}{k}.$

Adopt the movement velocity that can obtain vehicle after above-mentioned steps, operation complete once 1. ~ 4. after step, can repetitive operation 5. ~ 8. step, the unlimited measurement speed of a motor vehicle.Measure the speed of a motor vehicle as track need be changed, need first again operation steps 1. ~ 4., initialization is carried out to the data of new road, thus ensures the accuracy of vehicle speed measurement.

Claims (10)

1., based on a vehicle speed detection method for traffic surveillance videos, it is characterized in that, comprise the steps:

1. track segmentation;

2. lane width information is obtained;

3. the actual range representated by each pixel is calculated;

4. correction factor β is solved, in order to correct the actual track distance representated by each pixel;

5. vehicle movement track is obtained;

6. vehicle actual motion distance is calculated;

7. distance correction;

8. speed calculates.

2. the vehicle speed detection method based on traffic surveillance videos according to claim 1, is characterized in that, when step 1. split by track, adopts track marking line method of identification Delete superfluous background, obtains track picture; Original image element used during sectional drawing is consistent with video pixel.

3. the vehicle speed detection method based on traffic surveillance videos according to claim 1, is characterized in that, when 2. step carries out track pixel wide information acquisition, its lane width pixel number of lane width in video image represents.

4. the vehicle speed detection method based on traffic surveillance videos according to claim 1, is characterized in that, when 3. step calculates the track actual range representated by every pixel, supposes that the longitudinal width of each pixel is consistent with the distance representated by transverse width.

5. the vehicle speed detection method based on traffic surveillance videos according to claim 1, it is characterized in that, 4. step solves correction factor β, first distance calibration is carried out to track, thereafter compare actual range and calculate gained distance, obtain ratio between the two, as correction factor, in order to correct the actual track distance representated by each pixel; Two mark spacings of the actual range/calculating gained namely between β=two mark.

6. the vehicle speed detection method based on traffic surveillance videos according to claim 1, is characterized in that, when 5. step obtains vehicle movement track, supposes that vehicle does rectilinear motion.

7. the vehicle speed detection method based on traffic surveillance videos according to claim 6, is characterized in that, when 6. step calculates vehicle actual motion distance, distance solution formula sum formula replaces integral formula; Namely use the actual range sum in vehicle movement track representated by each pixel as the actual motion distance of vehicle.

8. the vehicle speed detection method based on traffic surveillance videos according to claim 1, it is characterized in that, when 6. step calculates vehicle actual motion distance, if the traveling of vehicle exists the change in direction, then need the deviation angle considering vehicle, in order to correct the actual motion distance of vehicle.

9. the vehicle speed detection method based on traffic surveillance videos according to claim 1, it is characterized in that, step is distance correction 7., is vehicle actual motion distance is multiplied by correction factor β.

10. the vehicle speed detection method based on traffic surveillance videos according to claim 1, is characterized in that, step 8. speed calculates, and is the actual motion speed calculating vehicle according to the move distance of vehicle in reality and run duration, namely wherein D be step 7. in the vehicle movement distance that calculates, T is front and back two frame video image frame difference/video frame rates.