CN100435160C - A video image processing method and system for real-time collection of traffic information - Google Patents

Abstract

The provided new detection technology with visual image for vehicle matching and traffic information acquisition comprises: pre-preparing image with background difference method, dynamic projecting to detect vehicle and extract vehicle feature information for vehicle matching by last and current frames of image and the corresponding traffic information. This invention can reduce computation quantity and improves reliability of result.

Description

A video image processing method and system for real-time collection of traffic information

technical field

The invention belongs to the technical field of intelligent traffic, and relates to traffic information collection technology.

Background technique

There are many technologies for collecting traffic information. Compared with other detection technologies, video detection technology has obvious advantages in coverage, detection parameters, maintainability, and ease of installation. It is forward-looking and can represent traffic information detectors. development trend.

At present, the principle of using video images to collect traffic information can be roughly divided into two categories: virtual coil method and tracking method. Most of the products on the market belong to the category of the virtual coil method. Its basic principle is to judge whether there is a vehicle passing by according to whether the gray level of the pixel at a fixed section of the road on the image changes, so as to count the traffic flow and calculate the speed. The advantage of this type of detector is that the principle is simple, the data processing time is short, and the flow and speed detection can be completed under the premise of meeting the real-time requirements. However, since this type of detector only obtains the only information of whether there is a vehicle passing the sampling line position at the sampling time, and loses features such as vehicle length, width, and motion trajectory, the vehicle matching accuracy of the front and rear frame images is low. In the case of overtaking and lane changes, vehicle matching errors lead to speed detection failures.

The principle of the tracking method is to identify the pixels in the traffic scene image that match the characteristics of the vehicle, count the number of vehicles, and match the vehicles in the front and rear frames according to the extracted features, so as to calculate the speed. In theory, the tracking method is more rigorous than the virtual coil method, so it can better represent the development trend. The difficulty of this method lies in: feature extraction and feature tracking. First of all, the features must be representative, and the vehicles in the image all have this feature and are different from each other; secondly, the features of the same vehicle in different frame images should be correlated and have a better correspondence. The disadvantage of the tracking method reported in the literature is that the calculation of the vehicle detection is too large and the result is not accurate.

Due to the complexity of obtaining traffic information by using video images, video acquisition technology is still in continuous improvement.

Contents of the invention

Aiming at the problems of large amount of calculation and unreliable results of the existing tracking method vehicle detection technology, the present invention provides a new vehicle detection technology, and based on this detection technology, it performs vehicle matching in front and rear frame images, thereby realizing information collection .

The processing flow of the present invention is shown in Figure 3, after setting specific parameters (comprising brightness, contrast, various thresholds), obtain a frame of video image, adopt dynamic projection method to detect vehicle, and then obtain the image of each vehicle on this frame image Information such as length, width, and shape is matched with the vehicle information of the previous frame image to realize the function of collecting traffic flow information.

Vehicle detection is how to distinguish whether the pixels on the image belong to the vehicle or the background, and it is necessary to identify whether different vehicle pixels belong to the same vehicle. This technology uses the dynamic projection method to detect vehicles. The process of this method is: firstly, perform difference processing with the background image, then perform horizontal projection, eliminate noise interference, perform image recognition, and extract vehicle feature information.

The vehicle matching process uses the vehicle information (including vehicle length, width and shape features, etc.) obtained after processing by the dynamic projection method to match with the image information of the previous frame.

Including the following steps:

(1) Threshold and image acquisition parameters in the detection process are set;

(2) The detection area is set by the control device, and multiple lanes can be operated in parallel at the same time;

(3) Use the background difference method to binarize the detection area. If the grayscale difference between the pixel and the background is greater than a certain threshold, it is considered to be the target pixel, and its grayscale value is assigned to a certain value and certain color to obtain a binary image;

(4) project the binary image in the horizontal direction to obtain a projection image;

(5) labeling is carried out to Fig. 2, and the continuous row of this color pixel is marked as the same numbering;

(6) Calculate the number of pixels of a certain color for each number and the distance between adjacent labels. When the number of target pixels is lower than the set threshold, or the distance between adjacent labels is lower than the set threshold, it is considered noise interference Exclude and adjust the label;

(7) take the last label as the total vehicle number of this frame image;

(8) Extract the feature of each label pixel block, including length, width and shape feature etc., as vehicle feature;

(9) Match with the vehicle characteristics of the previous frame image; if a vehicle in this frame image fails to match successfully, the total flow will increase by 1; if the match is successful, calculate the moving distance of the vehicle, calculate the vehicle speed, and calculate other traffic parameters ;

(10) Read the next frame image, go to step (3).

A video image processing system for real-time collection of traffic information, which has a structure for realizing any one of the above methods. It can be: the camera head used to collect signals is connected to a video recorder or an image recorder, and then connected to a computer through an image acquisition card and a digital/analog conversion device. The computer is provided with traffic information collection software; the video recorder The camera or image recorder can be connected with the image monitor at the same time.

Description of drawings

Fig. 1 is a binary graph of the present invention.

Figure 2 is a projected view of the present invention.

Fig. 3 is a schematic diagram of the system processing flow.

Fig. 4 is the setting parameter interface of the present invention.

Fig. 5 is a schematic diagram of the detection area setting interface of the present invention.

Fig. 6 is the length calibration interface (for calculating distance) of the present invention.

Fig. 7 is the detection operation interface of the present invention.

Fig. 8 is a schematic diagram of hardware connection of the present invention.

Detailed ways

Projection method and background difference method are a kind of technical means of single image recognition often used in image processing engineering. Projection method projects image pixels (usually 0-1 value image) in horizontal or vertical direction, and background difference method The method subtracts the pixel values of the image containing the object from the image not containing the object. The present invention adopts a new vehicle detection idea—dynamic projection method. This method comprehensively uses the projection method and the background difference method to continuously and dynamically process multiple images, and detects the vehicle in a single image by recognizing the projected image, and completes the vehicle detection in the front and rear frame images according to the characteristics of the vehicle projection image. Matching, so as to realize information collection.

Its basic principle is shown in Figure 1 and Figure 2 (white in the figure represents the target pixel), and the specific steps are as follows:

1. Set the threshold and image acquisition parameters (brightness and contrast) in the detection process;

2. Set the detection area through the control device (such as a mouse), and multiple lanes can be operated in parallel at the same time;

3. Use the background difference method to binarize the detection area (the pixel and background grayscale difference is greater than a certain threshold, then it is considered to be the target pixel, and its grayscale value is assigned to 255, i.e. white), to obtain Figure 1;

4. Project Figure 1 in the horizontal direction to obtain Figure 2;

5. Carry out label processing to Fig. 2, the continuous row of white pixel is marked as same numbering;

6. Calculate the number of white pixels for each number and the distance between adjacent labels. When the number of target pixels is lower than the set threshold, or the distance between adjacent labels is lower than the set threshold, it is considered to be noise interference and eliminated. Adjust the label;

7. Use the last label as the total number of vehicles in this frame image;

8. Extract the features of each labeled pixel block, including length, width and shape features, as vehicle features;

9. Match with the vehicle characteristics of the previous frame image. If a vehicle in this frame image fails to match successfully, the total flow will increase by 1; if the match is successful, calculate the moving distance of the vehicle, calculate the speed of the vehicle, and calculate other traffic parameters;

10. Read the next frame image and go to step 3;

Fig. 4-7 is the interface schematic diagram of specific embodiment; Fig. 8 is the hardware connection schematic diagram of a kind of embodiment of the present invention: the camera head 1 that is used to collect signal links to each other with video recorder or image recorder 2, then through Fig. Image acquisition card 3, digital/analog conversion device 4 are connected with computer 5, which is provided with traffic information collection software 6; this video recorder or image recorder 2 can be connected with image monitor 7 simultaneously.

Claims (5)

1. A video image processing method for real-time collection of traffic information, characterized in that: after the preprocessing of the image with the background difference method, the dynamic projection method is used to detect the vehicle, on this basis, the vehicle feature information is extracted, and the front and rear The vehicle matching of the frame image realizes the collection of traffic information. 2. The video image processing method for real-time collection of traffic information according to claim 1, characterized in that: the process of vehicle detection is: firstly perform difference processing with the background image, and then perform horizontal projection to eliminate noise interference, image recognition, and feature information extraction of vehicles; The vehicle matching process is to match the vehicle information obtained after processing with the dynamic projection method with the image information of the previous frame. 3. The video image processing method for real-time collection of traffic information according to claim 1 or 2, characterized in that it comprises the steps of: (1) setting thresholds and image acquisition parameters in the detection process; (2) The detection area is set by the control device, and multiple lanes can be operated in parallel at the same time; (3) Use the background difference method to binarize the detection area. If the grayscale difference between the pixel and the background is greater than a certain threshold, it is considered to be the target pixel, and its grayscale value is assigned to a certain value and certain color to obtain a binary image; (4) project the binary image in the horizontal direction to obtain a projection image; (5) labeling is carried out to this projection figure, and the continuous row of this color pixel is marked as the same numbering; (6) Calculate the number of pixels of a certain color for each number and the distance between adjacent labels. When the number of target pixels is lower than the set threshold, or the distance between adjacent labels is lower than the set threshold, it is considered noise interference Exclude and adjust the label; (7) take the last label as the total vehicle number of this frame image; (8) Extract the feature of each label pixel block, including length, width and shape feature, as vehicle feature; (9) Match with the vehicle characteristics of the previous frame image; if a vehicle in this frame image fails to match successfully, the total flow will increase by 1; if the match is successful, calculate the moving distance of the vehicle, calculate the vehicle speed, and calculate other traffic parameters ; (10) Read the next frame image, go to step (3). 4. The video image processing method for real-time collection of traffic information according to claim 3, characterized in that in step (3), the gray value is assigned 255, that is, white. 5. A system for video image processing for real-time collection of traffic information, characterized in that: it has a structure for realizing the method described in any one of claims 1-4: a camera and a video recorder for collecting signals Or the image recorder is connected, and then connected with the computer through the image acquisition card and the digital/analog conversion device, the computer is equipped with traffic information collection software; the video recorder or image recorder can be connected with the image monitor at the same time .

Images