KR19980074792A - Traffic information detection method and apparatus using image processing - Google Patents

Abstract

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a traffic information detecting method and apparatus using image processing. In the conventional algorithm approach for obtaining traffic information, the detection area installed in one lane must be sequentially detected when the vehicle passes, and thus the density of the vehicle. Is high, i.e. when a vehicle exists in more than one area at the same time, an error occurs, and if the distance between the two areas is narrowed to prevent the error, an error of calculation occurs, and the algorithm must separate the surface of the road from the vehicle. When the background image is continuously updated, the error occurs because the shape of the road is invisible due to the narrow head gap, and the pixel value of the road in the case of a vehicle having a contrast similar to the image of the road is assumed. Detection error occurs because the difference is small, and because the vehicle has bright pixel value, Many lice is detected an error in the case, the error also occurs to recognize the shadow caused by the vehicle in the lane next to the car, at night there is a problem that is further deepening the detection error.

Accordingly, the present invention has been made to solve the above-mentioned problems, and the image sensor is installed on the side of the road, and the image processing to detect the necessary traffic information by image processing the image of the road input through the image sensor. By providing the traffic information detection method and apparatus used, it is possible to connect with various devices for traffic control by using the image processing algorithm dualized during the day and at night, thereby improving the reliability of the traffic information detection.

Description

Traffic information detection method and apparatus using image processing

The present invention relates to a method and apparatus for detecting traffic information using image processing. In particular, an image sensor is installed on a roadside, and image processing of an image of a road input through the image sensor is used to detect necessary traffic information. It relates to a traffic information detection method and apparatus.

1 is a block diagram of a conventional image detecting apparatus for detecting traffic information. As shown in FIG. 1, a road that exists in a three-dimensional space is formed on a two-dimensional sensor so that detection is sequentially transmitted as an analog signal. A sensor unit 10; 3D space road formed by the analog / digital converter 12 for converting each pixel value of the image signal transmitted from the sensing sensor unit 10 into a digital value and transmitting the same, and a sensing sensor of the sensing sensor unit 10. An image acquisition unit 11 including a memory 13 for storing data corresponding to the image in a 1: 1 manner; A central processing unit 14 for controlling all functions of the image detecting apparatus; An operation process of the conventional device configured as the communication connection unit 15 for connecting with an external device will be described.

In the sensor 10, an image existing in a three-dimensional space is formed on a two-dimensional sensor, and the image is sequentially transmitted as an analog signal to the image acquisition unit 11, and the image acquisition unit 11 is used. The analog signal (video signal) transmitted to the digital signal is converted into a digital value of each pixel value of the input image through the analog / digital converter 12 in the image acquisition unit 11, and the values are converted into the image acquisition unit ( 11 is stored in the memory 13 in the memory 13, and the digital values sequentially stored in the memory 13 correspond to the road image of the three-dimensional space 1: 1, which is formed on the detection sensor of the sensor 10, Although there is a difference depending on the processing method, in general, the values of all the pixels stored in the memory 13 are not applied to the image processing algorithm, but only a portion determined at the time of initial setup is applied to the image processing algorithm, resulting in the burden of calculation. There is a difference in the detection area designation method at the time of initial setup according to the image processing algorithm.

Two basic algorithmic approaches are used to obtain traffic information, which is divided into Type 1 and Type 2 as follows.

<Type 1>

FIG. 2 illustrates an exemplary embodiment of positioning a virtual detection area on a conventional screen. As shown in FIG. 2, virtual detection areas 20 and 21 that are vertical to each other in which a user knows a distance from each other at the time of initial setup are shown on the screen. If there is no vehicle in the detection area (pixel array of M x N), the pixel values corresponding to each pixel are almost constant.

3 is an enlarged view of a detection area when a vehicle is in a conventionally designated detection area. As shown in FIG. 3, when the vehicle enters the detection area, the pixel portion 31 occupied by the vehicle and the pixel portion not occupied are shown. The pixel value changes, and the change determines whether the vehicle exists. If the vehicle proceeds downward, the two detection areas sequentially change the pixel value in the detection area and occupy until they pass through the detection area. Traffic information is calculated using the time difference between the one time and the two detection areas causing the pixel value change.

<Type 2>

The vehicle tracking algorithm detects moving blocks and measures the speed while following the trajectory.It uses only the information of a part of the image area to reduce the amount of calculation according to the complexity of the algorithm, and analyzes the entire image area and the vehicle in any direction. It is divided into ways to track it.

As described above, in the conventional algorithm approach for obtaining traffic information, the detection area installed in one lane must be detected sequentially when the vehicle passes. Therefore, when the vehicle density is high, that is, the vehicle exists in two or more areas at the same time. In this case, an error occurs, and if the gap between two areas is narrowed to prevent the error, a calculation error occurs, and the algorithm assumes that the surface of the road should be separated from the vehicle, and continuously updates the background image. If it is included, an error occurs because the shape of the road is invisible due to a narrow head gap, and a vehicle having a contrast similar to the image of the road generates a detection error because the difference between the pixel value of the road is small and the vehicle is bright. Detection error occurs even if the pixel value of the road is different from the pixel value of the road. Also it occurs due to the failure to recognize the shadow of the vehicle occurs, and at night there is a problem that is further deepening the detection error.

Accordingly, the present invention has been made to solve the above-mentioned problems, and the image sensor is installed on the side of the road, and the image processing to detect the necessary traffic information by image processing the image of the road input through the image sensor. An object of the present invention is to provide a method and apparatus for detecting traffic information.

1 is a block diagram of an image detecting apparatus for detecting conventional traffic information.

2 is an embodiment of designating a position of a virtual detection area on a conventional screen.

Fig. 3 is an enlarged view of the detection area when the vehicle is in the conventionally designated detection area.

4 is an operation flowchart of a traffic information detection apparatus using the image processing of the present invention;

Fig. 5 is an operation flowchart for changing the detection area setting.

6 is an embodiment of a method for designating a basic detection area for detecting traffic information on a road image.

7 is an operation flowchart of updating a reference image.

8 is an embodiment of a virtual detection model in the detection area for obtaining traffic information.

FIG. 9 is a pixel array in which the pixel value is changed when a vehicle exists in FIG. 8; FIG.

10 is a flowchart for calculating pixel values of the virtual detection model of FIG.

Fig. 11 is a flow chart of operation until one vehicle enters and passes the virtual detection area.

12 is a flowchart for calculating traffic information in accordance with a change of state.

Fig. 13A is a histogram of pixel values when the vehicle passes the virtual detection model.

Fig. 13B is a histogram of pixel values when the pixel value of the shadow is similar to the pixel value of the road side;

Fig. 13C is a histogram of pixel values when the pixel values are slightly different from the road surface.

Fig. 14 is a flowchart of discrimination between a shadow and a vehicle using the characteristic that pixel values of the shadow are concentrated in a narrow section of the histogram.

Fig. 15 is a pixel arrangement of the virtual detection model through which a vehicle passes at night.

16 is a vehicle detection flow chart using the characteristics of forming a blob.

Fig. 17 is a pixel array type diagram of a virtual detection model when a vehicle passes at night.

*** Description of the symbols for the main parts of the drawings ***

10: detection sensor unit 11: image acquisition unit

12: analog / digital converter 13: memory

14: central processing unit 15: communication connection

20, 21: Virtual detection area 22: 3 × 3 pixel array

P (I, J): pixel 31: pixel portion occupied by the vehicle

60a, 60b, 60c: basic detection area 61a, 61b: area of virtual detection model

62: guideline 80: first virtual detection area

81: second virtual detection area 82: third virtual detection area

90: fourth virtual detection area L: width of part occupied by vehicle

R: Width of the part occupied by the vehicle W: Threshold

130: portion of the headlight light reflected on the road 131: blob of the headlight light

132: left side of the virtual detection model 133: right side of the virtual detection model

The configuration of the present invention for achieving the above object comprises a sensing means for detecting the image of the road in the image sensor installed on the roadside through which the vehicle passes; Image acquisition and processing means for storing the image data sensed by the detection means in a memory and calculating traffic information; A central processing unit which manages traffic information calculated by the image acquisition and processing means and controls access to the outside; A communication connection unit for connecting to an external device; It consists of the operation means which sets a detection area and initializes an apparatus at the time of initial setup.

In addition, the method of the present invention includes a first step of determining whether it is necessary to correct or add a detection area designated to an image of a road provided by each image sensor during an initial setup of the device; A second step of resetting the detection area by the determination of the first step; A third step of determining whether or not the reference image, which is a road image without a vehicle, is updated; A fourth step of acquiring a new reference image by the determination of the third step; The fifth step of obtaining the current image from the image sensor and calculating the traffic information.

4 is an operation flowchart of the traffic information detection apparatus using the image processing of the present invention, and FIG. 5 is an operation flowchart for changing the detection area setting. As shown therein, a road provided by each image sensor in the initial setup process is shown. If the external device or the administrator needs to modify or add the designated detection area, and the communication device needs to change the detection area, proceed with setting the detection area, which is the change mode, and return to the work progress mode. 6 is an embodiment of a method for designating a basic detection area for detecting traffic information on an image of a road. As shown in FIG. 6, the detection area is first set by moving an operation means to the top and bottom of a lane boundary position. Specifying points (63, 64) in order creates a guideline (62), and in the case of multiple lanes, The guide and by specifying the starting and ending position of the line 62, between the pair of guide line to display the default detection area (60a, 60b, 60c).

If there is no change request for the detection area, it is immediately determined whether the base image, which is the work progress mode, is updated, and the road image continuously changes according to the time zone and weather changes. 7 is an operation flowchart of updating a reference image. As shown in FIG. 7, when a new image is input, the count is initialized to '0' and the pixel values of the basic detection regions 60a, 60b, and 60c of FIG. Sampling is performed to determine whether a vehicle exists by performing the following operation on the pixel array of the basic detection area, and each area is composed of the pixel array (M × N) specified in FIG. 2, and P (I, J) If is the pixel value,

--------------------------- (Equation 1)

The average value of the basic detection areas 60a, 60b, and 60c is obtained by using Equation 1, and the change in pixel value in the basic detection area is detected by calculating between the next Laplace operator and each pixel array.

Where L is the Laplace operator, I is an array of 3x3 pixel values centered on pixels [P (I, J)], and the result of the multiplication (LxI) of the Laplace operator and the pixel value array is The same applies to all pixel values in the basic detection areas 60a, 60b, 60c.

--- (Equation 2)

If the equation 2 is compared with the result of the previous image and the count is the same, the count is continuously increased, and since the vehicle does not exist in the basic detection areas 60a, 60b, and 60c, the count value is determined by the user. If the value is equal to or greater than the threshold, the image is acquired and used as the reference image. The traffic information calculation process of calculating the traffic volume, vehicle passing speed, average speed, occupancy rate, and vehicle type is performed using the reference image.

FIG. 8 is an embodiment of a virtual detection model in a detection area for obtaining traffic information. As shown in FIG. 8, it is detected whether a vehicle occupies the first, second, and third virtual detection models 80, 81, and 82. A state 1 in which no vehicle exists in the first, second, and third virtual detection models 80, 81, and 82; A state 2 in which the vehicle exists in the second virtual detection model 81 and no vehicle exists in the third virtual detection model 82; The vehicle passes through the second virtual detection model 81 and is classified into three states, which are the states 3 occupying the third virtual detection model 82. The presence or absence of the vehicle is largely divided into day and night. Judge.

FIG. 9 is a pixel array in which a pixel value is changed when a vehicle exists in FIG. 8, and FIG. 10 is a flowchart for calculating pixel values of the virtual detection model of FIG. 9, wherein the time zone is daytime. In this case, the images of the first, second, and third virtual detection models 80, 81, and 82 are acquired and compared with the pixel values at the same position among the pixels of the reference image, and if there is a difference, it is determined that the vehicle exists. By comparing the pixels on the right side of the second virtual detection model 81 occupied by the vehicle with the reference image, the width R of the occupied portion is obtained, and the pixels on the left side are occupied by comparing with the reference image. The width L of the portion that is present is obtained, and it is determined whether the sum R + L occupies the virtual detection model by comparing the threshold W with the size.

If a vehicle does not exist in the second virtual detection model 81, the vehicle may exist in the third virtual detection model 82. Therefore, the width R of the portion occupied by comparing the pixels on the right side with the reference image is compared. If the width R is less than the threshold W, the pixels on the left side are again compared with the reference image to obtain the width L of the occupied portion, and the width L is greater than the threshold W. If it is small, it is determined that there is no vehicle in the third virtual detection model 82, and it is determined whether the detected object is a vehicle or a shadow by a vehicle passing through the side lane.

FIG. 11 is a flowchart of operation until a vehicle enters and passes a virtual detection area, and FIG. 12 is a flowchart for calculating traffic information according to a change of state. As shown in FIG. The process of 10 is applied to each frame of the image which is continuously input, and as a result, a combination of the three states is made, and the current time measurement is continued to calculate the period to update the reference image, and after a certain time The process of updating the reference image is performed through the process of FIG. 7 by determining the reference image update. When the current state is the state 1 by determining the state of the reference image, it is determined whether the vehicle exists in the second virtual detection model 81. If is present, the time (T1) the vehicle entered into the virtual detection model 81, the state is switched to state 2 to determine the presence of the vehicle, if the vehicle exists 3 records the virtual detection model 82 time (T2), the vehicle is entering the calculated speed ( , S: actual distance), and when the vehicle exists, it is determined whether the vehicle exists at the same time. The fourth virtual detection model 90 of FIG. 9 is automatically generated, and the determination of the vehicle model is performed by the first virtual detection model 80. If the vehicle enters at the same time and the vehicle does not exist in the second virtual detection model 81, the small car, and if the vehicle exists only in the first and second virtual detection models 81 and 81, then the first, second and third virtual cars If present in the detection model (80, 81, 82) is classified as a large vehicle, and if the vehicle does not exist in the third virtual detection model (82), it is switched to the state 3 to determine whether the vehicle has passed and determine the passing time of the vehicle Record and calculate the time and occupancy time it takes for the vehicle to pass between the first and second virtual detection models 81 and 82, calculate the average speed (moving average) of the recently passed vehicles, It is calculated by (H: head interval time).

Fig. 13A is a histogram of pixel values when the vehicle passes the virtual detection model, and Fig. 13B is a histogram of pixel values when the pixel value of the shadow is similar to the pixel value of the roadside. Pixel value histogram when there is a difference. As shown in this figure, the range of pixel values is divided into N sections, and the cumulative values of the pixel values included in each section are shown. Since most pixel values of shadows have similar pixel values, they tend to be concentrated in a narrow section, and these characteristics are used to discriminate between shadows and vehicles.

14 is a determination flowchart for using a characteristic pixel values of the shadow to be concentrated in a narrow interval of the histogram shadow vehicle, equal parts N the range of the pixel value to the d ₁ ~ d _N as shown to this, and a virtual Detection Model Determining the section to which the pixel values belong, and when the degree of each section is X ₁ to X _N , the values are calculated to determine the section (d _M ) with the maximum value among the number of values, and the second highest number determining the interval _(S d) having to calculate the histogram to determine the amount of (C) to produce a shadow by the equation 3 below.

--------------- (Equation 3)

Epsilon (3) ) Is a constant used to normalize the histogram discrimination amount (C), and is a constant that changes according to the environment of the road. do.

FIG. 15 is a pixel array of a virtual detection model in which a vehicle passes at night. As shown in FIG. 15, when the time zone is at night, the headlights of a vehicle entering the vehicle are reflected on a road, and thus the first, second, and third virtual detection models (80). , 81, 82 change the pixel values very irregularly, but the central part of the headlamp has the highest pixel value, and the pixel values around the headlamp centering the headlamp form one blob 131. Have

FIG. 16 is a flowchart of a vehicle detection using a characteristic of forming a blob, and FIG. 17 is a pixel array type diagram of a virtual detection model when a vehicle passes at night, and the left side of the virtual detection model when the vehicle enters the virtual detection model. A type 1 for detecting a part, detecting a right side of the virtual detection model when a blob of a headlight is found, and ending detection of a vehicle when a light of a headlight is found; Detect the left side of the virtual detection model, detect the right side of the virtual detection model if the blob of the headlight is found, detect the left side of the virtual detection model again if the light of the headlight is not found, and if the light of the headlight is found, Type 2 and ending index finger; Detect the left side of the virtual detection model, detect the right side of the virtual detection model again if the headlight is not found, detect the right side of the virtual detection model if the blob of the headlight is detected, and detect the vehicle's detection if the blob is detected. It is divided into Type 3, which continues the detection of the vehicle through this process, and proceeds with the traffic information detection process based on the result.

As described above, the traffic information detection method and apparatus using the image processing of the present invention can be connected to various devices for traffic control by using the image processing algorithm dualized during the day and night, and improve the reliability of the traffic information detection. It is effective to let.

Claims (8)

Sensing means for detecting an image of a road by an image sensor installed on a roadside through which a vehicle passes; Image acquisition and processing means for storing the image data sensed by the detection means in a memory and calculating traffic information; A central processing unit which manages traffic information calculated by the image acquisition and processing means and controls access to the outside; A communication connection unit for connecting to an external device; A traffic information detecting device using image processing, characterized by comprising operation means for setting a detection area and initializing the device during initial setup. A first step of determining whether a detection area setting change is necessary to correct or add a detection area designated to an image of a road provided by each image sensor during an initial setup of the device; A second step of resetting the detection area by the determination of the first step; A third step of determining whether or not the reference image, which is a road image without a vehicle, is updated; A fourth step of acquiring a new reference image by the determination of the third step; And a fifth step of obtaining the current image from the image sensor and calculating the traffic information. The method of claim 2, wherein the first step comprises: allocating initialization data to a memory and determining whether a change request for a detection area is requested by an external device or an administrator; And, if there is a change request for the detection area, proceeding to set the detection area in the change mode, and if not, proceeding to the work progress mode. According to claim 2, wherein the second step by using the operation means on the screen of the management computer to create a boundary line by designating a point set at the top and bottom of the boundary position of the lane, automatically in the area between the two boundary lines A traffic information detection method using image processing, comprising: setting a virtual detection area. 3. The method of claim 2, wherein the third step comprises: initializing a count to '0' when a new image is input, and sampling pixel values of the basic detection area; Determining whether a vehicle does not exist in the basic detection area; And determining whether the same image has been continued for a predetermined time or more, and obtaining a reference image. The method of claim 2, wherein the fourth step is occupied by comparing pixels in the right portion of the virtual detection model occupied by the vehicle with the reference image to determine whether the vehicle exists in the virtual detection model when the time zone is daytime. Find the width (R) of the part, compare the pixels in the left part with the reference image, get the width (L) of the part occupied, and compare the sum with the magnitude of the threshold to determine whether the virtual detection model is occupied. Making a step; If the vehicle does not exist in the virtual detection model, the vehicle may exist in another virtual detection model. Therefore, the width of the occupied area is obtained by comparing the pixels on the right side with the reference image. Comparing the pixels in the image to a reference image to obtain a width of the occupied portion, and determining that there is no vehicle in the virtual detection model when the width is smaller than the threshold value; And determining whether the detected object is a vehicle or a shadow by a vehicle passing through a side lane. The method of claim 2, wherein the fourth step detects the left part of the virtual detection model when the vehicle enters the virtual detection model when the time zone is night, and detects the right side of the virtual detection model when the blob of the headlight is detected. Ending detection of the vehicle when the blob of the headlight is detected; Detect the left part of the virtual detection model, and if the blob of the headlight is detected, detect the right side of the virtual detection model.If the blob of the headlight is not detected, detect the left part of the virtual detection model again. Ending the index finger; The left side of the virtual detection model is detected, and if the blob of the headlight is not detected, the right side of the virtual detection model is detected again. If the blob of the headlight is detected, the right side of the virtual detection model is detected and the blob of the vehicle is detected. Traffic information detection method using the image processing, characterized in that it comprises a step. The method of claim 2, wherein the fifth step comprises: dividing a range of pixel values into intervals to use characteristics of a histogram; Determining intervals to which the pixel values of the virtual detection model belong, to obtain the degree values of each interval; And determining a histogram discriminating amount using the number of taps in each section to discriminate between a vehicle and a shadow.