CN102136196A - Vehicle velocity measurement method based on image characteristics - Google Patents

Abstract

The invention discloses a method for measuring vehicle speed based on a video sequence. The method obtains image information of the lane by installing a fixed camera directly above the lane, and sets two virtual coils A and B on the lane image, and detects the coil through continuous multi-frames. The change of the image feature information in the vehicle captures whether the vehicle has entered the coil, and when the set threshold condition is met, the signal is output and the time is recorded, so as to obtain the time difference between the vehicle entering coil A and B and the real distance between the two coils Calculate the speed of the vehicle. The invention is simple, real-time and efficient for vehicle speed measurement, has strong robustness to changes in illumination conditions and slight camera shakes in the real-time environment, and can effectively suppress the influence of shadows on detection.

Description

A vehicle speed measurement method based on image features

technical field

The invention relates to the fields of image processing and intelligent transportation, in particular to a vehicle speed measurement method based on image features.

Background technique

Vehicle speed measurement is an important part of intelligent transportation system, which provides real-time traffic data for traffic supervision and management, and vehicle speed measurement is carried out on the basis of vehicle detection. There are usually two types of vehicle detection methods: intrusive and non-invasive. The former is inconvenient to install, such as loop coil detection, piezoelectric sensors, magnetic sensors, pneumatic road tubes, etc. These devices need to be installed directly on the road when they are used. Or buried under the road surface; the latter is easy to install, non-destructive, and cost-saving, such as video image processing, microwave radar, infrared sensors, ultrasonic detectors, etc. Among them, the advantages of video image processing are the most prominent, easy to install and can It can be completed by adding corresponding functional modules on the basis of the existing monitoring system, which greatly saves expenses. It also has traffic supervision and traffic management functions, and can provide visual images for traffic management departments. Video-based vehicle detection methods are divided into two categories: the first category is based on virtual points, virtual lines, and virtual coils; the second category is based on target extraction and tracking, and they have their own advantages and disadvantages.

Although the vehicle speed measurement method based on target extraction and tracking has a large detection coverage area, its algorithm is relatively complex and its real-time performance is poor. The method based on virtual coil is simple and effective, but the general vehicle speed measurement method based on virtual coil is too simple, and it is not robust to light changes, especially susceptible to shadows. In addition, due to the influence of the environment, the camera will shake slightly in real-time applications, which will bring large errors to the detection.

Contents of the invention

In view of this, the main purpose of the present invention is to provide a vehicle speed measurement method based on image features. The invention realizes the acquisition of the running speed of the vehicle appearing in the image through the video image processing technology, which is simple, real-time and efficient, and has strong robustness to environmental changes.

In order to achieve the above object, the present invention provides a method for measuring vehicle speed based on image features, comprising the steps of:

A. Set two virtual coils on the lane image, calculate the real distance between the coils, and obtain sub-image areas of corresponding sizes;

B. Perform image processing and analysis on the sub-image area, capture the signal when the vehicle passes through the two coils and obtain the time;

C. Calculate the traveling speed of the vehicle according to the signal and time, and the real distance.

Described step A comprises:

A1. Collect real-time images of the lane;

A2. On the acquired image containing the lane, set two virtual coils according to the lane;

A3, converting the image coordinates to real world coordinates, thereby calculating the real distance between the two coils;

A4. According to the image coordinate position of the virtual coil on the original image, intercept the image area at the corresponding position to generate a sub-image, thereby completing the downsampling of the original image and simplifying subsequent processing.

Described step B comprises:

B1. Calculate the currently acquired image feature information of the sub-image area to obtain a feature image;

B2. Perform pixel-level and block-level subtraction operations on the feature image and the pre-established image feature background model, and synthesize the results of the two types of subtraction operations according to the judgment criteria to obtain a new image feature difference image;

B3. Updating the image feature background model;

B4. Make statistics on the image features of the foreground image, judge whether it satisfies the condition for the vehicle to enter the virtual coil, and if so, output a signal that the vehicle enters the coil and record the current time.

Wherein, the step of setting up the image feature background model described in step B2 includes:

Obtain a sequence of image frame data, and calculate its corresponding image feature information;

A statistical model is established and initialized by using the feature information as a basic unit of image processing, so as to obtain a background reference image based on image features.

Wherein, the step of performing a block-level subtraction operation on the feature image and the image feature background model described in step B2 includes:

Perform the following operations on the feature image and the image feature background model respectively: take a certain pixel as the center, perform cumulative statistics on the values of the pixels within its domain, and use the statistical value as the value of the current pixel. value, and so on, until the entire image is traversed to generate a new feature image;

Subtracting the two generated feature images to obtain a difference image.

Wherein, the step of synthesizing the results of the two types of subtraction operations described in step B2 to obtain a new image feature difference image according to the judgment criterion includes:

First, take out a certain pixel in the differential image obtained by the pixel-level operation;

Then fetching a pixel corresponding to the pixel position in the differential image obtained by the block-level operation;

The two pixels with the same position are judged as follows: if the value of the pixel of the pixel-level difference image is not empty, and the value of the pixel of the block-level difference image is empty, then the The value of the corresponding pixel position on the new image feature difference image is set to be non-null, and finally a new difference image is obtained after traversing the entire image.

Wherein, the step of judging whether the condition for the vehicle to enter the virtual coil described in step B4 includes:

Perform statistics on the feature information of the foreground image to obtain a feature description about the foreground image;

If the feature description satisfies the condition for judging whether an image area has changed, it indicates that the image area corresponding to the coil has changed, and the change is counted;

It is judged whether the counting times meet the threshold condition when the vehicle enters the coil, if so, it is judged that the change of the image area is caused by the vehicle entering the coil, then output the signal of the vehicle entering the coil and record the current time.

Described step C comprises:

C1, judging whether the second virtual coil generates a signal, if so, continue to step C2;

C2. Judging whether the first virtual coil has generated a signal, if so, marking the two signals as signals generated when the vehicle passes through the two coils, otherwise invalid;

C3. Obtain the time difference when the vehicle passes the two coils according to the time corresponding to the two signals;

C4. According to the real distance between the two coils obtained in the step A, calculate the driving speed of the vehicle through a speed calculation formula.

The beneficial effects of the present invention are:

A vehicle speed measurement method based on image features proposed by the present invention obtains the driving speed of the vehicle appearing in the image through video image processing technology, which is simple, real-time and efficient, and has strong robustness to illumination changes and slight camera shakes , which can effectively suppress the influence of shadows on detection.

Description of drawings

Fig. 1 is a flow chart of an embodiment of the vehicle speed measurement method based on image features of the present invention;

Fig. 2 is a flow chart of an embodiment of the image preprocessing module in Fig. 1;

Fig. 3 is a schematic diagram of an embodiment of setting a virtual coil in the present invention;

Fig. 4 is a flow chart of an embodiment of the vehicle capture module in Fig. 1;

Fig. 5 is a flowchart of an embodiment of block-level edge feature background subtraction in Fig. 4;

FIG. 6 is a flowchart of an embodiment of the vehicle speed calculation module in FIG. 1 .

Detailed ways

The present invention will be further described in detail below through specific embodiments in conjunction with the accompanying drawings.

Please refer to Figure 1, a vehicle speed measurement method based on image features includes steps:

S1, the image preprocessing module, that is, two virtual coils are set on the lane image, the real distance between the coils is calculated, and a sub-image area of a corresponding size is obtained;

S2. The vehicle capture module, which performs image processing and analysis on the sub-image area, captures the signal when the vehicle passes through the two coils and obtains the time;

S3. The vehicle speed calculation module, that is, calculate the driving speed of the vehicle based on the signal, time, and the real distance.

Please refer to Fig. 2, in one embodiment of the present invention, step S1 comprises the steps:

S11. Setting virtual coils A and B. In this embodiment, the virtual coil adopts a rectangular shape, as shown in FIG. 3 , of course, other irregular shapes can also be adopted;

S12. Calculate the real distance between the two coils, first obtain the pixel coordinates of the coils on the image, then convert them into the world coordinate system, and then obtain the real coordinates of a point at the corresponding positions of the two coils, so that Calculate the real distance between the two coils;

S13. According to the image coordinate position of the virtual coil on the original image, intercept the image area at the corresponding position to generate a sub-image, thereby completing the downsampling of the original image and simplifying subsequent processing.

S14. Output the sub-image data.

Please refer to Fig. 4, in one embodiment of the present invention, step S2 comprises the steps:

S21. Calculate the edge feature of the sub-image to obtain its edge feature image;

S22. Perform binarization and denoising processing on the edge feature image;

S23. Perform a pixel-level image subtraction operation on the edge feature image of the sub-image and the edge feature background model, so as to obtain an edge feature difference image. The pixel-level operation refers to using a single pixel as a basic processing unit;

S24. Perform a block-level image subtraction operation on the edge feature image of the sub-image and the edge feature background model, so as to obtain an edge feature difference image;

Please refer to FIG. 5. In this embodiment, the block-level subtraction operation of the feature image and the image feature background model includes steps:

S241. Generate a new edge feature image according to the edge feature image of the sub-image. In this embodiment, the following method is adopted: taking a certain pixel point as the center, performing cumulative statistics on the values of the pixel points within its range, using the statistical value as the value of the current pixel point, and so on, Until the entire image is traversed, a new edge feature image is generated. In this embodiment, the field adopts a rectangle whose length and width are 3 pixels in size, of course, fields of different sizes and shapes can also be used; there are only two types of pixel values in the edge feature image: 1 or 0, and of course Other values can be taken;

S242. Generate a new edge feature background reference image according to the edge feature background reference image. In this embodiment, the following method is adopted: taking a certain pixel point as the center, performing cumulative statistics on the values of the pixel points within its range, using the statistical value as the value of the current pixel point, and so on, Until the entire image is traversed, a new edge feature background reference image is generated. Wherein, the domain type and size remain consistent with the domain used in step S241. In this embodiment, there are only two pixel values of the edge feature background reference image: 1 or 0, and of course other values can also be used;

S243. Perform a subtraction operation on the two generated edge feature images to obtain a difference image;

S244. Perform image binarization and denoising processing on the difference image;

S245. Output the edge feature difference image processed in step S244.

S25, according to the difference image obtained respectively in step S23 and step S24, carry out the slight shake elimination processing of the camera, including steps:

First take out a certain pixel in the differential image obtained in step S23;

Then take out the pixel corresponding to the difference image obtained in the step S24 on the pixel position;

The two pixels are judged as follows: if the value of the pixel of the pixel-level difference image is not empty, and the value of the pixel of the block-level difference image is empty, then the new image The value of the corresponding pixel position on the feature difference image is set to be non-null, and finally a new difference image is obtained after traversing the entire image. In this embodiment, a pixel value of 1 means not empty, and 0 means empty.

S26. Update the established background model of the edge feature image according to the currently obtained frame information of the binary difference image. In this embodiment, the updating method may adopt a moving average method, and of course an average or median method may also be used;

S27. Count the total number of foreground points of the characteristic image. In this embodiment, the foreground points can be set as pixel points with a value of 1, and of course other values can also be used.

S28 , judging whether the obtained total number of foreground points is greater than the set threshold T1 , if yes, execute step S29 , otherwise execute step S210 . In this embodiment, the threshold T1 is a percentage, that is, the ratio of the total number of foreground points to the total number of pixels in the coil. The threshold T1 can be set according to experience.

S29. Increase the confidence degree CAR IN of the vehicle corresponding to the current coil entering it by 1;

S210. Decrease the confidence degree CAR IN of the vehicle corresponding to the current coil entering it by 1;

S211. Judging whether the confidence degree CAR IN corresponding to the coil is greater than the set threshold T2, if yes, execute step S212. The threshold T2 can be set according to experience.

S212. Output a signal that the vehicle enters the coil, and record the current time Time.

Please refer to FIG. 6, in one embodiment of the present invention, step S3 includes steps:

S31. Judging whether the second virtual coil B has a signal generated, if so, proceed to step S32;

S32. Judging whether the first virtual coil A has a signal generated, if so, proceed to step S33;

S33. Obtain the time Time corresponding to the two signals, so as to obtain the time difference when the vehicle passes the two coils;

S34. Obtain a real distance between the two coils;

S35. According to the distance and the time difference, obtain the traveling speed of the vehicle through a speed calculation formula.

S36. Output the driving speed of the vehicle.

The above content is a further detailed description of the present invention in conjunction with specific embodiments, and it cannot be assumed that the specific implementation of the present invention is limited to these descriptions. For those of ordinary skill in the technical field of the present invention, without departing from the concept of the present invention, some simple deduction or replacement can be made, which should be regarded as belonging to the protection scope of the present invention.

Claims (8)

1. A method for measuring vehicle speed based on image features, comprising the steps of: A. Set two virtual coils on the lane image, calculate the real distance between the coils, and obtain sub-image areas of corresponding sizes; B. Perform image processing and analysis on the sub-image area, capture the signal when the vehicle passes through the two coils and obtain the time; C. Calculate the traveling speed of the vehicle according to the signal and time, and the real distance. 2. The method according to claim 1, wherein said step A comprises: On the acquired image containing the lane, set two virtual coils according to the lane; The real distance between the two coils is obtained through calibration and conversion calculation between the image coordinates and the real world coordinates; The image area corresponding to the image coordinate position of the coil is intercepted to generate a sub-image, thereby completing the down-sampling of the original image. 3. The method according to claim 1, wherein said step B comprises: Calculate the image feature information of the currently acquired sub-image area to obtain the feature image; Perform pixel-level and block-level subtraction operations on the feature image and the pre-established image feature background model, and synthesize the results of the two types of subtraction operations according to the judgment criteria to obtain a new image feature difference image; updating the image feature background model using the foreground image; The image features of the foreground image are counted, and it is judged whether it satisfies the condition of the vehicle entering the virtual coil, and if so, a signal of the vehicle entering the coil is output and the current time is recorded. 4. The method according to claim 3, wherein the initialization and establishment of the image feature background model comprises: acquiring a sequence of image frame data, calculating its corresponding image feature information, using the feature information as The basic unit of image processing establishes and initializes a statistical model to obtain a background reference image based on image features. 5. The method according to claim 3, wherein said performing a block-level subtraction operation on the feature image and the image feature background model comprises: Perform the following operations on the feature image and the image feature background model respectively: take a certain pixel as the center, perform cumulative statistics on the values of the pixels within its domain, and use the statistical value as the value of the current pixel. value, and so on, until the entire image is traversed to generate a new feature image; Subtracting the two generated feature images to obtain a difference image. 6. method as claimed in claim 3 is characterized in that, described according to the result of described two kinds of deduction operation comprehensively according to judgment criterion and obtains new image characteristic difference image and comprises: First, take out a certain pixel in the differential image obtained by the pixel-level operation; Then fetching a pixel corresponding to the pixel position in the differential image obtained by the block-level operation; The two pixels with the same position are judged as follows: if the value of the pixel of the pixel-level difference image is not empty, and the value of the pixel of the block-level difference image is empty, then the The value of the corresponding pixel position on the new image feature difference image is set to be non-null, and finally a new difference image is obtained after traversing the entire image. 7. The method according to claim 3, wherein the judging whether the condition for the vehicle entering the virtual coil is satisfied comprises: Perform statistics on the feature information of the foreground image to obtain a feature description about the foreground image; If the feature description satisfies the condition for judging whether an image area has changed, it indicates that the image area corresponding to the coil has changed, and the change is counted; It is judged whether the counting times meet the threshold condition when the vehicle enters the coil, if so, it is judged that the change of the image area is caused by the vehicle entering the coil, then output the signal of the vehicle entering the coil and record the current time. 8. The method according to claim 1, wherein said step C comprises: When the second virtual coil generates a signal, it is judged whether the first virtual coil has output a signal, and if so, the two signals are marked as the signals respectively generated when the vehicle passes through the two coils , otherwise invalid; Obtaining the time difference when the vehicle passes the two coils according to the time corresponding to the two signals; Further, according to the real distance between the two coils obtained in the step A, the driving speed of the vehicle is calculated through a speed calculation formula.

Images