CN107194386A - A kind of intersection electric bicycle travel speed acquisition methods based on video - Google Patents

Abstract

The invention discloses a video-based method for acquiring the driving speed of an electric bicycle at a road intersection, comprising the following steps: 1. Collecting a vehicle driving video at a road intersection, and setting the detection area in the video as an area surrounded by multiple sidewalks; 2. 1. Process the collected video, identify and track the electric bicycle at the intersection of the road in the video, and obtain the coordinate points of the driving trajectory; 3. Convert the centroid coordinates of the electric bicycle on the video image into actual two-dimensional plane coordinates, and calculate its driving speed. This method can accurately obtain the driving speed of the electric bicycle inside the road intersection.

Description

A video-based method for obtaining the speed of electric bicycles at road intersections

technical field

The invention belongs to the field of traffic monitoring, and in particular relates to a method for obtaining the running speed of an electric bicycle by using computer vision technology.

Background technique

In recent years, the proportion of electric bicycles in the transportation structure has been increasing. Compared with walking, electric bicycles are more energy-saving and faster; compared with cars, they have the advantages of electric drive, no emissions, environmental friendliness, less pollution, no fear of congestion and low cost. Electric bicycles, a fast, clean and low-cost way of travel, are very popular among people, and have gradually become the main mode of transportation for short-distance travel in cities.

However, while the travel mode of electric bicycles is rapidly developing and popularizing in my country, it also brings problems such as congestion and safety to urban road intersections. Electric bicycle riders disregard road safety laws and regulations, such as electric bicycles often entering the driving space of motor vehicles at intersections, violating traffic signals, speeding and rushing, and electric vehicle owners modifying electric vehicles for cargo transportation without permission, which is the cause of electric bicycles. An important factor of low driving safety. How to reasonably analyze and standardize the driving behavior of electric bicycles at intersections requires obtaining the driving parameters of electric bicycles, including driving speed, safe driving area and safe driving speed.

At present, the extraction of micro-traffic data of electric bicycles at home and abroad is mainly based on fuzzy mathematical calculations. This investigation method has a lot to do with people's subjective initiative. The investigators' observation of the data is unstable, and the calculated micro-traffic data is often The precision is lower.

Contents of the invention

Purpose of the invention: Aiming at the problems existing in the prior art, the present invention provides a video-based method for obtaining the speed of electric bicycles at road intersections, which can accurately obtain the speed of electric bicycles inside road intersections.

Technical solution: The present invention adopts the following technical solution: a video-based method for obtaining the driving speed of an electric bicycle at a road intersection, comprising the following steps:

(1) Collect the vehicle driving video at the road intersection, and set the detection area in the video as the area surrounded by multiple sidewalks;

(2) Process the collected video, identify and track the electric bicycle at the road intersection in the video, and obtain the coordinate points of the driving track;

(3) Convert the coordinates of the center of mass of the electric bicycle on the video image to the actual two-dimensional plane coordinates, and calculate its driving speed.

Step (2) comprises the following steps:

(2-1) set up the background image model f (x, y) of the collected road intersection video; Can adopt the Gaussian mixture background modeling method to set up the background image model f (x, y);

(2-2) Make a difference between the original image of the tth frame in the video and the background image model established in step (2-1), and obtain the foreground moving target A(x _t , y _t ) of the tth frame; the calculation formula of the foreground moving target is: :

A(x _t ,y _t )=f(x _t ,y _t )-f(x,y)

Among them, f(x _t , y _t ) is the original image of frame t, and f(x, y) is the background image;

(2-3) select multiple frames of images in the video, and the foreground moving target in each frame of images is the foreground blob of an electric bicycle, calculate its blob area, the length of the blob circumscribed rectangle, the width of the blob circumscribed rectangle;

(2-4) set up electric bicycle parameter model, the parameter of described model is electric bicycle agglomerate area S, the length L of agglomerate circumscribed rectangle, the width W of agglomerate circumscribed rectangle; To the parameter that obtains in (2-3) Carry out statistics, select the parameter range that distribution probability is η as the value range of electric bicycle parameter model;

(2-5) For the i-th frame image and the j-th frame image in the video, obtain the foreground moving objects A _i and A _j , calculate the corresponding foreground blob area, the length of the blob circumscribed rectangle, and the blob circumscribed rectangle width, According to the value range of parameters in the electric bicycle parameter model, judge whether A _i and A _j are electric bicycles;

(2-6) If both A _i and A _j are electric bicycles, calculate the centroids p _i and p _j of A _i and A _j , and the centroid distance d _ij , if the conditions are met: Then A _i and A _j are the same electric bicycle, and p _i and p _j are the driving trajectory coordinate points of the electric bicycle in the i-th frame image and the j-th frame image.

Step (3) comprises the following steps:

(3-1) The actual driving distance d _ab of the electric bicycle in the first frame image and the second frame image is:

Where (u _a , v _a ) is the two-dimensional ground coordinates of the electric bicycle in the a-th frame image; (u _b , v _b ) is the two-dimensional ground coordinates of the electric bicycle in the b-th frame image;

(3-2) The average speed v _ab of the electric bicycle between the first frame image and the second frame image is:

Where t _ab is the time difference between the a-th frame image and the b-th frame image.

The instantaneous speed Speed _f of the image in frame f of the electric bicycle is:

Where r is the frame rate of the video, (u _f , v _f ) is the two-dimensional ground coordinates of the electric bicycle in the fth frame image, (u _f+1 , v _f+1 ) is the electric bicycle in the f+1th frame image The two-dimensional ground coordinates of .

Beneficial effects: Compared with the prior art, the video-based method for obtaining the driving speed of electric bicycles at road intersections disclosed by the present invention has the following advantages: 1. It can accurately obtain the driving trajectory and driving speed of electric bicycles inside the intersection; 2. 1. The video sources taken at different intersections have the excellent characteristics of repeatable detection methods, high accuracy and easy modification of parameters, which is of great significance to the analysis of driving behavior of electric bicycles at intersections.

Description of drawings

Fig. 1 is the flowchart of the method provided by the present invention;

Figure 2 is a detailed flow chart of the electric bicycle tracking steps;

Fig. 3 is an original image and a background image detection area diagram;

Fig. 4 is a comparison diagram of the foreground image before and after noise reduction of the foreground image;

Fig. 5 is the effect drawing of electric bicycle screening;

Fig. 6 is a diagram of the safe driving range of electric vehicles at road intersections;

Fig. 7 is the hourly distribution diagram of the maximum speed of electric bicycles at road intersections.

detailed description

In this embodiment, the acquisition of the driving speed of an electric bicycle at the intersection of Zhongshan East Road-Taiping North Road in Nanjing is taken as an example to further illustrate the present invention.

As shown in Figure 1, the present invention provides a kind of video-based method for obtaining the traveling speed of an electric bicycle at a road intersection, comprising the following steps:

(1) Collect the vehicle driving video at the road intersection, and set the detection area in the video as the area surrounded by multiple sidewalks;

Shoot the Zhongshan East Road-Taiping North Road intersection from a top-down perspective; set the detection area 1 as the area surrounded by multiple sidewalks at the intersection, such as the polygonal area in Figure 3(b).

(2) process the collected video, identify and track the electric bicycle at the road intersection in the video, and obtain the driving trajectory coordinate points; as shown in Figure 2, specifically include the following steps:

(2-1) set up the background image model f (x, y) of the collected road intersection video; The present embodiment adopts the mixed Gaussian background modeling method to set up the background image model f (x, y);

(2-2) Make a difference between the original image of the tth frame in the video and the background image model established in step (2-1), and obtain the foreground moving target A(x _t , y _t ) of the tth frame; the calculation formula of the foreground moving target is: :

A(x _t ,y _t )=f(x _t ,y _t )-f(x,y)

Among them, f(x _t , y _t ) is the original image of frame t, and f(x, y) is the background image;

After obtaining the foreground moving target A(x _t , y _t ), filter operation is performed on A(x _t , y _t ) to remove the shadow part and noise of the electric bicycle. In this embodiment, A(x _t , y _t ) is filtered Gaussian filtering, the filtering effect is shown in Figure 4, where (a) is the foreground image before filtering; (b) is the foreground image after filtering.

(2-3) select multiple frames of images in the video, and the foreground moving target in each frame of images is the foreground blob of an electric bicycle, calculate its blob area, the length of the blob circumscribed rectangle, the width of the blob circumscribed rectangle;

(2-4) set up electric bicycle parameter model, the parameter of described model is electric bicycle agglomerate area S, the length L of agglomerate circumscribed rectangle, the width W of agglomerate circumscribed rectangle; To the step (2-3) that obtains The parameters are counted, and the parameter value range with a distribution probability of 80% is selected as the value range of the electric bicycle parameter model;

(2-5) For the i-th frame image and the j-th frame image in the video, obtain the foreground moving objects A _i and A _j , calculate the corresponding foreground blob area, the length of the blob circumscribed rectangle, and the blob circumscribed rectangle width, According to the value range of the parameters in the electric bicycle parameter model, judge whether A _i and A _j are electric bicycles; as shown in Figure 5, the box in the figure is the foreground image judged to be electric bicycles.

(2-6) If both A _i and A _j are electric bicycles, calculate the centroids p _i and p _j of A _i and A _j , and the centroid distance d _ij , if the electric bicycles in the two frames of images are the same, then their The distance traveled in a very short time will not exceed a very small value, so set the frame difference threshold f _th to control the time, and set the distance threshold d _th to control the distance. In this embodiment, d _th is set to 20 pixels, and f _th is set to 20 frames. If the conditions are met: Then A _i and A _j are the same electric bicycle, and p _i and p _j are the driving trajectory coordinate points of the electric bicycle in the i-th frame image and the j-th frame image.

The formula for calculating the centroid is:

Among them, m _pq is the p+q order moment of the image, (x _c , y _c ) is the centroid of the image, A(x, y) is the detected electric bicycle clump image, N and M are the length and width of the image respectively value.

The formula for calculating the distance between the mass centroids of the i-th frame image and the j-th frame image is:

Where (x _ci , y _ci ) is the coordinates of the center of mass of the electric bicycle clump in the i-th frame image.

(3) Convert the coordinates of the center of mass of the electric bicycle on the video image to the actual two-dimensional plane coordinates, and calculate its driving speed.

In this embodiment, the homogeneous coordinate conversion method is used to convert the pixel coordinates into two-dimensional ground coordinates. The error analysis between the conversion distance and the actual distance is shown in Table 1:

Table 1

Across the four samples, all error rates for the tests were below 5%, with an average error rate of 3%. The intersection is 20 meters long, and the total error of the electric bicycle driving through the intersection is 60 cm, which can be ignored for the conflict analysis of electric bicycles.

The steps to calculate the average speed of the electric bicycle between two frames of images are:

(3-1) The actual driving distance d _ab of the electric bicycle in the first frame image and the second frame image is:

Where (u _a , v _a ) is the two-dimensional ground coordinates of the electric bicycle in the a-th frame image; (u _b , v _b ) is the two-dimensional ground coordinates of the electric bicycle in the b-th frame image;

(3-2) The average speed v _ab of the electric bicycle between the first frame image and the second frame image is:

Where t _ab is the time difference between the a-th frame image and the b-th frame image.

The instantaneous speed Speed _f of the image in frame f of the electric bicycle is:

Where r is the frame rate of the video, (u _f , v _f ) is the two-dimensional ground coordinates of the electric bicycle in the fth frame image, (u _f+1 , v _f+1 ) is the electric bicycle in the f+1th frame image The two-dimensional ground coordinates of .

The safe driving area and safe driving speed in the road intersection can be obtained by making statistics on the driving speed of electric bicycles passing through the road intersection. The steps are as follows:

Statize the maximum speed of electric bicycles passing through road intersections. The maximum driving speed stipulated by the state is 20km/s, that is, 5.5m/s. For electric bicycles that are not speeding, obtain the cumulative trajectory of their driving trajectories, and draw a map of the safe driving area inside the intersection. As shown in Figure 6.

The maximum speed of electric bicycles passing through road intersections in peak hours is counted. The distribution of the maximum speed is shown in Table 2, and the distribution of the maximum speed is shown in Figure 7.

Table 2

From Table 2 and Figure 7, it can be seen that the maximum speed of electric bicycles is distributed in the four intervals above 5.5-6.5m/s, 6.5-7.5m/s, 7.5-8.5m/s, and 8.5m/s. For electric vehicles, the distribution probabilities are 5%, 4%, 2% and 0% respectively, and the overall distribution probability of the maximum speed of the electric vehicles driving over the speed limit is 11%.

The maximum speed distribution of electric bicycles without speeding is most concentrated between 2.5-3.5m/s, with a probability of 46%. Arrange the non-speeding speed data from small to large, remove the first 15% of the speed data and the last 15% of the speed data, and select the remaining speed parameter distribution interval as the safe driving speed interval [2.5, 5.0], the unit is m/s .

Claims (5)

1. A video-based method for obtaining the driving speed of an electric bicycle at a road intersection, is characterized in that, comprising the steps: (1) Collect the vehicle driving video at the road intersection, and set the detection area in the video as the area surrounded by multiple sidewalks; (2) Process the collected video, identify and track the electric bicycle at the road intersection in the video, and obtain the coordinate points of the driving track; (3) Convert the coordinates of the center of mass of the electric bicycle on the video image to the actual two-dimensional plane coordinates, and calculate its driving speed. 2. the video-based road crossing electric bicycle travel speed acquisition method according to claim 1, is characterized in that, step (2) comprises the steps: (2-1) set up the background image model f (x, y) of the road crossing video that is collected; (2-2) Make a difference between the original image of the tth frame in the video and the background image model established in step (2-1), and obtain the foreground moving target A(x _t , y _t ) of the tth frame; the calculation formula of the foreground moving target is: : A(x _t ,y _t )=f(x _t ,y _t )-f(x,y) Among them, f(x _t , y _t ) is the original image of frame t, and f(x, y) is the background image; (2-3) select multiple frames of images in the video, and the foreground moving target in each frame of images is the foreground blob of an electric bicycle, calculate its blob area, the length of the blob circumscribed rectangle, the width of the blob circumscribed rectangle; (2-4) set up electric bicycle parameter model, the parameter of described model is electric bicycle agglomerate area S, the length L of agglomerate circumscribed rectangle, the width W of agglomerate circumscribed rectangle; To the parameter that obtains in (2-3) Carry out statistics, select the parameter range that distribution probability is η as the value range of electric bicycle parameter model; (2-5) For the i-th frame image and the j-th frame image in the video, obtain the foreground moving objects A _i and A _j , calculate the corresponding foreground blob area, the length of the blob circumscribed rectangle, and the blob circumscribed rectangle width, According to the value range of parameters in the electric bicycle parameter model, judge whether A _i and A _j are electric bicycles; (2-6) If both A _i and A _j are electric bicycles, calculate the centroids p _i and p _j of A _i and A _j , and the centroid distance d _ij , if the conditions are met: Then A _i and A _j are the same electric bicycle, and p _i and p _j are the driving trajectory coordinate points of the electric bicycle in the i-th frame image and the j-th frame image. 3. the video-based road crossing electric bicycle traveling speed acquisition method according to claim 1, is characterized in that, step (3) comprises the steps: (3-1) The actual driving distance d _ab of the electric bicycle in the first frame image and the second frame image is: <mrow> <msub> <mi>d</mi> <mrow> <mi>a</mi> <mi>b</mi> </mrow> </msub> <mo>=</mo> <msqrt> <mrow> <msup> <mrow> <mo>(</mo> <msub> <mi>u</mi> <mi>a</mi> </msub> <mo>-</mo> <msub> <mi>u</mi> <mi>b</mi> </msub> <mo>)</mo> </mrow> <mn>2</mn> </msup> <mo>+</mo> <msup> <mrow> <mo>(</mo> <msub> <mi>v</mi> <mi>a</mi> </msub> <mo>-</mo> <msub> <mi>v</mi> <mi>b</mi> </msub> <mo>)</mo> </mrow> <mn>2</mn> </msup> </mrow> </msqrt> </mrow> Where (u _a , v _a ) is the two-dimensional ground coordinates of the electric bicycle in the a-th frame image; (u _b , v _b ) is the two-dimensional ground coordinates of the electric bicycle in the b-th frame image; (3-2) The average speed v _ab of the electric bicycle between the first frame image and the second frame image is: <mrow> <msub> <mi>v</mi> <mrow> <mi>a</mi> <mi>b</mi> </mrow> </msub> <mo>=</mo> <mfrac> <msub> <mi>d</mi> <mrow> <mi>a</mi> <mi>b</mi> </mrow> </msub> <msub> <mi>t</mi> <mrow> <mi>a</mi> <mi>b</mi> </mrow> </msub> </mfrac> </mrow> Where t _ab is the time difference between the a-th frame image and the b-th frame image. 4. the method for obtaining the speed of electric bicycles at road intersections based on video according to claim 1, wherein the instantaneous speed Speed _f of electric bicycles at the f frame image is: <mrow> <msub> <mi>Speed</mi> <mi>f</mi> </msub> <mo>=</mo> <mi>r</mi> <msqrt> <mrow> <msup> <mrow> <mo>(</mo> <msub> <mi>u</mi> <mi>f</mi> </msub> <mo>-</mo> <msub> <mi>u</mi> <mrow> <mi>f</mi> <mo>+</mo> <mn>1</mn> </mrow> </msub> <mo>)</mo> </mrow> <mn>2</mn> </msup> <mo>+</mo> <msup> <mrow> <mo>(</mo> <msub> <mi>v</mi> <mi>f</mi> </msub> <mo>-</mo> <msub> <mi>v</mi> <mrow> <mi>f</mi> <mo>+</mo> <mn>1</mn> </mrow> </msub> <mo>)</mo> </mrow> <mn>2</mn> </msup> </mrow> </msqrt> </mrow> Where r is the frame rate of the video, (u _f , v _f ) is the two-dimensional ground coordinates of the electric bicycle in the fth frame image, (u _f+1 , v _f+1 ) is the electric bicycle in the f+1th frame image The two-dimensional ground coordinates of . 5. The method for obtaining the speed of electric bicycles at road intersections based on video according to claim 2, wherein step (2-1) adopts a mixed Gaussian background modeling method to establish a background image model f (x, y).