CN106951821A - A method for intelligent monitoring and recognition of smoky vehicles based on image processing technology - Google Patents

Abstract

The invention discloses a kind of black smoke car intelligent monitoring recognition methods based on image processing techniques, this method includes：Step 1: gathering traffic block port high definition real time video data stream using video frequency collection card；Step 2: the time sampled using changing distance background mean value Sampling techniques, the degree control interval changed according to road surface, recycles multiple image mean algorithm to extract road surface background information；Step 3: according to the video flowing of collection, filtering dilly and non power driven vehicle in real time using dolly filter algorithm to each two field picture, reducing the data volume of image procossing, improve efficiency of algorithm；Step 4: being positioned to the oversize vehicle afterbody smoke evacuation in each two field picture；Whether Step 5: the localization region image of exhaust emissions and background area image are carried out into Data Comparison, it is black smoke car to judge the vehicle.Black smoke car intelligent monitoring recognition methods disclosed in this invention can improve detection efficiency, judge that vehicle whether there is black smoke problem using computer technology.

Description

A method for intelligent monitoring and recognition of smoky vehicles based on image processing technology

technical field

The invention relates to the field of image pattern recognition and intelligent transportation, in particular to a method for identifying whether a vehicle emits black smoke by using image processing technology in an intelligent transportation system, especially for large-scale engineering vehicles and vehicles with excessive emissions.

Background technique

The pollution of smoky vehicles has always been the focus and difficulty of motor vehicle environmental protection. At present, the phenomenon of smoky emissions from diesel vehicles is still very serious. 80% of motor vehicle pollutants in a city are emitted by 20% of high-pollution vehicles, of which medium and heavy Diesel vehicles are big emitters. The pollutants emitted by a heavy-duty diesel vehicle are equivalent to the emissions of 500 small cars 1:300~500. A city that manages 10,000 heavy-duty diesel vehicles is equivalent to managing 3-5 million small cars. Therefore, it is an inevitable trend to study an image processing algorithm through machine vision to solve this problem.

There are mainly two traditional management methods for smoky vehicles, one is the manual road inspection mode, and the other is the manual viewing surveillance video mode. Although the two modes reduce the pollution of smoky vehicles to a certain extent, due to the With the sharp increase in the number of vehicles and the heavy traffic, the efficiency of road inspection and manual video viewing is not only inefficient, but also there are many difficulties. The environmental protection of motor vehicles urgently needs an online monitoring mode with a high degree of automation.

Contents of the invention

In order to overcome the above-mentioned problems and deficiencies of the prior art methods and realize efficient and intelligent automatic detection and identification of smoky vehicles, the present invention provides an intelligent monitoring and identification method for smoky vehicles based on image processing.

The technical scheme adopted in the present invention is:

Use high-definition cameras at traffic checkpoints to collect real-time video data streams;

Adopt variable interval background average sampling technology, adjust the interval sampling time parameter PTIME according to the degree of road surface change, take continuous 8-second video during sampling, and use multi-frame image average algorithm to extract road surface background information.

According to the collected video stream, the car filter algorithm is used to filter small vehicles and non-motorized vehicles in real time for each frame of image, reducing the amount of image processing data and improving the efficiency of the algorithm;

Detect whether there is a large vehicle in each frame of the filtered image, and if there is a large vehicle, locate its rear exhaust.

Compare the data of the positioning area image of the exhaust emission with the background area image to judge whether the vehicle is a smoky vehicle.

The technical effect of the present invention lies in: utilizing the high efficiency of the computer, designing an automatic black-smoky vehicle recognition technology, and judging whether the vehicle has black smoke problem according to the data analysis.

Description of drawings

Figure 1 is the effect of background extraction

Figure 2 is an explanatory analysis diagram of the frame difference method

Figure 3 is the effect map of black smoke positioning

Fig. 4 is an overall flowchart of the present invention.

detailed description

The present invention will be further described in detail below in conjunction with the accompanying drawings. Fig. 4 has shown the overall flowchart of the present invention.

As shown in FIG. 1 , the present invention is composed of four modules: a background extraction module S1 , a filter trolley module S2 , a tail smoke exhaust location module S3 , and a discrimination module S4 .

S1: First of all, for a section of traffic surveillance video, the first step is to extract its background information. The background is the road condition when there are no vehicles and other moving objects in the picture at that moment. Using the variable interval background sampling method, according to the change of the road surface, Adjust the interval of background sampling in real time. When there is a big change between the background of the previous sampling and the background of the current sampling, the time interval of the next sampling will be shortened, otherwise the current sampling interval will be maintained. Using variable interval sampling background can effectively adapt to weather changes and road surface changes, and adaptively adjust the sampling interval to maximize the efficiency and accuracy of the algorithm. Avoid that when the road surface conditions change, there will be a big discrepancy between the changed road surface and the collected background, which will seriously affect the subsequent processing results.

Using variable interval background sampling strategy can better adapt to complex background changes. The extracted background image is saved and used in the subsequent processing stages. The background extraction algorithm uses a multi-frame image mean algorithm to extract a video of about M seconds, with P frames of images per second, and uses formula 1 to calculate the average value of each pixel as the background pixel, where n=M*P is the total The picture frame numbers R, G, and B are the RGB channel values of the i -th picture frame respectively. Figure 1 The effect diagram of background extraction.

S2: Use the frame difference method to calculate the position of the vehicle and the approximate vehicle shape information. This method can only retain the trajectory of moving objects in three consecutive image frames, and remove the interference of unstable objects at the same time. Therefore, when the detected object changes greatly at different times, the frame difference method has better applicability. The specific principles are as follows:

Extract three consecutive frames of grayscale images from the video, record them as f _k-1 , f _k , and f _k+1 in sequence in time, and divide the current frame image f _k and its previous frame image f _k-1 according to formula 2, Make a difference and threshold to obtain a binary image d _k-1,k . Similarly, f _k and the next frame image f _k+1 are subtracted and thresholded to obtain a binary image d _k,k+1 . In the binary image d _k-1,k and d _k,k+1 , the region with pixel "0" represents the static background, and the region with pixel "1" represents the moving foreground. Then make the two binary images do logical "AND" operation, so as to extract the moving target area in the current frame f _k . Specific process

As shown in Figure 2.

After detecting the position and shape information of the vehicle, the connected area of the vehicle is processed, the coordinates of the vehicle and the number of vehicles are calculated, and the size of the vehicle is obtained by scanning and analyzing each connected area row by row, and the length of the vehicle is calculated. Remove small vehicles and irrelevant vehicles that are less than the preset parameters, eliminate interference and reduce the number of judgments on smoky vehicles.

S3: The tail smoke exhaust positioning module completes the positioning of the vehicle exhaust pipe. After the completion of S2, the position information of each large vehicle is obtained. In S3, the position information obtained by S2 will be used to obtain the DES of the vehicle rear. The distance from the left boundary of the vehicle LDES, and the distance from the right boundary of the vehicle RDES are smoke exhaust areas, and the values of DES, LDES, and RDES are related to the length and width of the vehicle and the position of the vehicle in the image.

S4: Discrimination module, this step will use the background information of S1, the vehicle location information of S2 and the exhaust position of the tail of S3, and the comprehensive results can be obtained. The specific steps are as follows:

When judging the smoky car of vehicle A, it is first judged whether the vehicle completely appears in the monitoring screen and is a certain pixel distance away from the left border and the right border of the screen.

The smoke exhaust location area at the rear of the vehicle is extracted with a rectangular frame, and the size of the rectangular frame is related to the length and width of the vehicle.

Calculate the value of the pixel points in the rectangular frame of the vehicle rear smoke exhaust positioning area and record it as Q1, take the same rectangular frame position in the background image and calculate the RGB value of the pixel point in the rectangular frame of the background image as Q2, Z=│Q1 -Q2│, segmented judgment is made according to the threshold of Z. When Z<10, it is judged as a non-smoky vehicle. When Z>20, it is judged as a smoky vehicle. When it is a non-smoky vehicle, no output is made. In the case of a black-smoky vehicle, different black-smog levels are judged according to the Z value, and a screenshot of the vehicle’s black-smog emission is output and saved. The result is shown in Figure 3.

Claims (5)

1. An intelligent monitoring and identification method for smoky vehicles based on image processing technology. Using the high efficiency of computers, an automated smoky vehicle identification technology is designed to judge whether the vehicle emits black smoke based on data analysis. It is different from existing smoky vehicles. The identification method compared to its characteristics includes the following steps: Step 1. Use the video capture card to collect high-definition real-time video data streams at traffic checkpoints. The camera needs to be installed 5 meters above the road, and the shooting direction should not be blocked by foreign objects. turn-in section; Step 2. Adopt variable interval background average sampling technology, adjust the time parameter PTIME of interval sampling according to the degree of road surface change, take continuous 8-second video during sampling, and use multi-frame image average algorithm to extract road surface background information; Step 3. According to the collected video stream, use the car filtering algorithm to filter small vehicles and non-motorized vehicles in real time for each frame of image, reduce the amount of data for image processing, and improve the efficiency of the algorithm; Step 4. Locate the smoke exhaust from the rear of the large vehicle in each frame of image; Step 5: Compare the data of the positioning area image of the exhaust emission with the background area image to determine whether the vehicle is a smoky vehicle. 2. The method for intelligent monitoring and identification of smoky vehicles based on image processing technology according to claim 1, characterized in that said step 2 includes: Use the variable interval background sampling method to adjust the interval of background sampling in real time according to the change of the road surface. When the background of the previous sampling and the current sampling background have a large change, shorten the time interval of the next sampling, otherwise keep the current sampling interval; Using multi-frame image mean algorithm to extract road background information. 3. The method for intelligent monitoring and identification of smoky vehicles based on image processing technology according to claim 1, characterized in that the step 3 includes: Judging the size of the vehicle, when it is not smaller than the size of the target recognition vehicle, use the small car filtering algorithm to exclude the interfering vehicle. 4. The method for intelligent monitoring and identification of smoky vehicles based on image processing technology according to claim 1, characterized in that said step 4 includes: The rear smoke exhaust positioning module determines the parameters of DES, LDES, and RDES according to the vehicle's position and length and width information, and determines the coordinates of the vehicle's smoke exhaust area. 5. The method for intelligent monitoring and identification of smoky vehicles based on image processing technology according to claim 1, characterized in that the step five includes: Compare the data of the positioning area image of the exhaust emission with the background area image, and determine whether it is a smoky car by comparing the values.