KR20190076271A - Method and System for Analysis of Road Congestion robust to Vehicle Stagnation Using CCTV - Google Patents

Abstract

The present invention relates to a method and a system for analyzing road congestion robust to vehicle stagnation by using images from CCTVs. The method for analyzing road congestion robust to vehicle stagnation by using images from CCTVs comprises the steps of: allowing a camera unit to photograph roads; allowing a reference background extracting unit to extract a background from a captured image; allowing a road area extraction unit to extract the maximum area without edges from the extracted background and verify a road area by using an optical flow calculation result to extract the road area; allowing a foreground extracting unit to divide the extracted road area into quadrangular grids, extract the characteristics of each grid of a current frame, calculate differences between the road area and the characteristics of each grid of the current frame, and extract a foreground in which grids each having predetermined difference or more are recognized as a vehicle; and allowing a congestion analyzing unit to calculate the ratio of the foreground area to the extracted road area to analyze congestion. Thus, the foreground of the road can be stably extracted.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and system for analyzing road congestion using a CCTV image,

The present invention relates to a road congestion analysis method and system robust against vehicle congestion using a CCTV image. More particularly, the present invention relates to a method and an apparatus for analyzing a road congestion degree by detecting a background image and a foreground image stably when a vehicle is stationary, To a system and method for analyzing road congestion using a ratio.

Road congestion refers to the degree of occurrence of continuous delays and running costs of vehicles when the number of vehicles running on the road exceeds the limit, and in the past, most of the road congestion is observed and estimated by the loop detector. However, this point observation technology requires costly equipment and is difficult to install, and a method of analyzing the road congestion using the already installed CCTV is suggested.

It is important to extract the background and the foreground when analyzing the road congestion using CCTV images. In the conventional background extraction methods, the intensity values of the pixels in the image can be modeled using the Gaussian mixture model, There is a problem that the stagnation of the vehicle is recognized as a background and the accuracy is lowered.

A problem to be solved by the present invention is to provide a method and system for analyzing road congestion that is robust to vehicle congestion using CCTV image that reliably extracts the road foreground by using edge information despite the congestion of the road, .

According to an aspect of the present invention, there is provided a method of analyzing a road congestion that is robust against a vehicle congestion using a CCTV image. The method includes a step of photographing a road of a camera section, a step of extracting a background from the captured image, Extracting a maximum area without an edge in a background in which the road area extracting unit is extracted, extracting a road area after verifying the road area using the optical flow calculation result, Extracting a characteristic of each lattice of the current frame by dividing the characteristic of each lattice into a plurality of lattices, calculating characteristic differences of the lattice between the road region and the current frame, The congestion degree is calculated by calculating the ratio occupied by the foreground area in the road area from which the congestion analysis section is extracted Includes steps.

Through the above-mentioned problem solving means, the present invention can enhance the background extraction of the CCTV image, so that the road panorama can be stably found despite the road congestion.

Also, by extracting the background and the foreground using a more accurate background and foreground extraction method from the CCTV shot image installed on the road, the accuracy is higher than the conventional method.

In addition, cost reduction is achieved by analyzing the road congestion using CCTV, which is installed without any additional equipment.

1 is a flowchart illustrating a method for analyzing road congestion using CCTV according to an embodiment of the present invention.
2 is a block diagram illustrating a method of analyzing a road congestion degree analysis system using CCTV according to an embodiment of the present invention.
3 is a flowchart illustrating a method of extracting a background according to an embodiment of the present invention.
4 is a flowchart illustrating a method of extracting a road region according to an embodiment of the present invention.
5 is a flowchart illustrating a method of extracting a foreground according to an embodiment of the present invention.
6 is a flowchart illustrating a method of analyzing road congestion according to an exemplary embodiment of the present invention.
FIGS. 7 and 8 are illustrations for explaining a background and foreground extraction method according to an embodiment of the present invention.
9 is a flowchart illustrating an image processing method in a CCTV rotation according to an embodiment of the present invention.
10 and 11 are views illustrating an image processing method in a CCTV rotation according to an embodiment of the present invention.

It is to be understood that the specific structural or functional description of embodiments of the present invention disclosed herein is for illustrative purposes only and is not intended to limit the scope of the inventive concept But may be embodied in many different forms and is not limited to the embodiments set forth herein.

The embodiments according to the concept of the present invention can make various changes and can take various forms, so that the embodiments are illustrated in the drawings and described in detail herein. It is not intended to be exhaustive or to limit the invention to the particular forms disclosed, but on the contrary, is intended to cover all modifications, equivalents, or alternatives falling within the spirit and scope of the invention.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The singular expressions include plural expressions unless the context clearly dictates otherwise. In this specification, the terms "comprises" or "having" and the like are used to specify that there are features, numbers, steps, operations, elements, parts or combinations thereof described herein, But do not preclude the presence or addition of one or more other features, integers, steps, operations, components, parts, or combinations thereof.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings attached hereto.

1 is a flowchart illustrating a method of analyzing a road congestion that is robust against a vehicle congestion using a CCTV image according to an exemplary embodiment of the present invention. Referring to FIG. 1, the camera unit of the analysis system first photographs a road in real time (S110), and extracts a background based on the photographed image (S120). The road area is then extracted from the extracted background (S130), and the foreground is extracted from the road area (S140). The road congestion is analyzed based on the extracted foreground (S150).

2 is a block diagram illustrating a method of analyzing a road congestion degree analysis system using CCTV according to an embodiment of the present invention.

2, the road congestion analysis system 100 includes a camera unit 110, a reference background extraction unit 120, a vehicle region extraction unit 130, a road region extraction unit 140, a foreground extraction unit 150, A control unit 160, a storage unit 170, a communication unit 180, a rotation sensing unit 190, and a congestion analysis unit 195.

The camera unit 110 may photograph the road in real time and may be a CCTV already installed.

The reference background extraction unit 120 extracts a background using a MOG2 algorithm after capturing an input frame of an image photographed from the camera unit 110, and extracts a background of a minimum edge by measuring the number of edges of the background. The edge extraction algorithm is combined with the background extraction algorithm to extract the minimum edge background.

The road area extracting unit 130 extracts a maximum edgeless area from the background extracted from the reference background extracting unit 120, extracts a road area after verifying the road area using the optical flow calculation result .

The foreground extracting unit 140 extracts characteristics of each lattice by dividing the road region extracted from the road region extracting unit 130 into square lattices, extracts characteristics of each lattice of the current frame, And calculates the difference of the characteristic, and extracts the foreground which recognizes the lattice over a certain difference as the vehicle.

The control unit 150 controls the process of the process related to the operation of the system, and controls the operation of each configuration. The storage unit 160 may store images photographed by the camera unit, and may store the extracted background, road area, foreground, and congestion. The communication unit 170 is a communication module that exchanges data with an external device.

The rotation sensing unit 180 captures the next image and performs edge analysis. When the rotation is detected, the rotation sensing unit 180 performs the edge analysis again, compares the image with the previous image, and stores the image in the image queue. In particular, when the frequency of the transverse edge increases rapidly based on the occurrence frequency of the transverse edges, it is recognized as the rotation.

The congestion analysis unit 190 calculates a ratio occupied by the foreground region in the extracted road area to analyze the congestion degree. That is, the road area can be divided by the actual ratio to obtain the area ratio. That is, the area weight can be differentiated by distinguishing the trapezoidal area in the trapezoidal road area from the trapezoidal area in the proximity of the camera and the trapezoidal area in the far area. That is, the trapezoidal area close to the camera and the trapezoidal area spaced apart from the camera are distinguished from each other in the trapezoidal road area. The width of the adjacent trapezoidal area is broader than the width of the trapezoidal area spaced apart from the camera. The area weight of the vehicle detected in the spaced trapezoidal area is calculated to be higher, and the error according to the perspective of the vehicle detected in the area far from the camera can be reduced.

3 is a flowchart illustrating a method of extracting a background according to an embodiment of the present invention.

Referring to FIG. 3, an input frame of an image photographed from a camera unit is captured (S310), and a background is extracted using a MOG2 algorithm (S320). The MOG2 algorithm is a Gaussian of Mixture-based background / foreground segmentation algorithm, which can select the appropriate Gaussian distribution value for each pixel. The extracted background image is extracted by using a canny edge detector to measure the number of edges (S330). The background having the minimum edge is extracted (S340)

4 is a flowchart illustrating a method of extracting a road region according to an embodiment of the present invention.

Referring to FIG. 4, the background of the minimum edge is input (S410), and the maximum edgeless area is extracted from the background of the minimum edge (S420). Thereafter, the road area is verified using the optical flow calculation result (S430). The final road area is extracted according to the verification result (S440).

5 is a flowchart illustrating a method of extracting a foreground according to an embodiment of the present invention. Referring to FIG. 5, the extracted road area and current frame are inputted (S510), and the road area is divided into a square grid (S520). The property of each lattice of the road region is extracted (S530), and the property of each lattice of the current frame is extracted (S540). The difference between the characteristics of the road region and the current frame is calculated (S550), and the foreground image of the grid is recognized as a vehicle (S560).

6 is a flowchart illustrating a method of analyzing road congestion according to an exemplary embodiment of the present invention.

Referring to FIG. 6, a road area and a foreground are input (S610). The ratio of the foreground area in the road area is calculated (S620). The road congestion degree using the area ratio is calculated (S630).

FIGS. 7 and 8 are illustrations for explaining a background and foreground extraction method according to an embodiment of the present invention. 7 (b) is an extracted background image according to the MOG2 background extraction algorithm, FIG. 7 (c) shows a frame of the minimum edge component from the extracted background image, Fig. That is, since the edge of the background image increases when the vehicle stagnates, the edge of the background image is minimized to extract only the road region. By combining these edge analysis algorithms, it is possible to solve the problem of residual image on the background image when congestion occurs.

FIG. 8A is an exemplary view showing a road area extracted by a background extraction algorithm and an edge analysis algorithm. FIG. 8B is an example of dividing a road area as an actual ratio in order to recover a situation that is ignored due to perspective to be. That is, the road area can be divided by the actual ratio to obtain the area ratio. That is, the area weight can be differentiated by distinguishing the trapezoidal area in the trapezoidal road area from the trapezoidal area in the proximity of the camera and the trapezoidal area in the far area. That is, the trapezoidal area close to the camera and the trapezoidal area spaced apart from the camera are distinguished from each other in the trapezoidal road area. The width of the adjacent trapezoidal area is broader than the width of the trapezoidal area spaced apart from the camera. The area weight of the vehicle detected in the spaced trapezoidal area is calculated to be higher, and the error according to the perspective of the vehicle detected in the area far from the camera can be reduced.

9 is a flowchart illustrating an image processing method in a CCTV rotation according to an embodiment of the present invention.

Referring to FIG. 9, an image processing method when a periodically rotating CCTV camera is used will be described. First, the image is photographed (S910), the edge is analyzed in the photographed image (S920), the rotation is sensed (S930), the additional image is processed using the image in the state where the rotation is stopped and the rotation is stopped (4S940). When the camera is stopped, it is judged by comparing the direction of the camera with respect to which direction the current camera faces. If an image having a high degree of matching is found (S960), the corresponding frame is registered in the image frame queue so that additional image processing can be performed (S970). If no matching image is found, discard the image and wait for the next rotation. It calculates the Y direction edge of Sobel Edge Detection with respect to the basic rotation detection and detects the rotation of the CCTV image when a sudden change occurs.

In the equations (1) and (2), I is the input image, Gx in the equation (1) is the result of applying the Sobel filter in the X direction, and Gy in the equation (2) is the result of applying the Y direction Sobel filter. All the matrix values of Equation (1) are set to 0 and only the Y direction component is used.

10 and 11 are views illustrating an image processing method in a CCTV rotation according to an embodiment of the present invention.

Referring to FIGS. 10 and 11, in FIG. 10 (a), the camera rotates horizontally due to the characteristics of the rotating CCTV. As a result, as shown in FIG. 10 (b) Can be seen. Because of this characteristic, the vertical edge is relatively sharply reduced.

Since the CCTV rotation detection algorithm uses only the simple features of the edge to detect it, it can rarely be perceived as a rotation even though the camera is not rotating, or it can be mistaken as if it was rotated twice. It is necessary to reconfirm the direction of the CCTV image by comparing the position of the CCTV image with the direction of the camera immediately after detecting the stop of the rotation in case that the CCTV rotation detection algorithm detects whether the rotation is uncorrected or unrecognized. 11 (a) is a result of histogram comparison of the HSV color with the A, B, and C images for the reference image on the upper left, respectively. In case of the same direction image, the color distribution is expected to be similar. However, as shown in the experimental results, the road like the reference is 11% even though it is B, the matching rate is very low. In order to solve this problem, the present invention proposes a divided histogram method in which images are finely divided and compared. 11 (b) is a result of applying the divided histogram algorithm based on the picture as shown in FIG. 11 (a). After applying the histogram comparison algorithm to each divided image, it selects the image with the highest matching rate.

Histogram result of segmented image Number recognized as part of A. Number recognized as part of B Number recognized as part of C 76 95 129

Table 1 shows the result of the divided histogram, and B should be the highest, but the number of C is the highest. After applying the algorithm for obtaining the road area part of the image, the result is shown in Fig. 11 (c), except for the road area part.

Histogram matching rate results excluding road area A B (actual match image) C Existing histogram comparison 18% 11% 45% Split histogram comparison 18% 61% 21%

As shown in Table 2, even if B shows 11% agreement with the existing histogram, the split histogram except the road region can achieve a high output of 61%. This figure shows that although the result obtained through the conventional histogram comparison method shows a match rate of 45% that does not match the actual road, the image of the actually matching B in the method of the present invention shows the highest match rate of 61% .

Rotation detection rate Detection speed Actual number of revolutions ratio 50 50 100%

Table 3 shows the result of executing the rotation detection for the image with the divided color histogram. It detects 100% of all rotations in one rotating image and showed 100% accuracy. That is, the present invention detects the rotation through the reduction of the vertical edges of the Sobel-edge detector, and after using the algorithm to obtain the road area, the divided histogram algorithm is applied to calculate the histogram value of the divided image in the background portion with the reference image The rotation correction is performed through comparison.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is evident that many alternatives, modifications and variations will be apparent to those skilled in the art. Accordingly, the true scope of the present invention should be determined by the technical idea of the appended claims.

100; Road Congestion Analysis System; 110; Camera part
120; A reference background extraction unit 130; The road area extracting unit
140; A foreground extracting unit 150; The control unit
160; Storage unit 170; Communication section
180; A rotation sensing unit 190; Congestion Analysis Unit

Claims (7)

The present invention relates to a method for analyzing road congestion that is robust against vehicle congestion using CCTV images,
Photographing a road section of a camera;
Extracting a background from the photographed image;
Extracting a maximum edgeless area in the background from which the road area extraction unit is extracted, extracting a road area after verifying the road area using the optical flow calculation result,
The road area from which the foreground extracting unit is extracted is divided into a square lattice to extract lattice-specific characteristics, and characteristics of each lattice of the current frame are extracted to calculate characteristic differences of the lattice between the road area and the current frame, Extracting the recognized foreground; And
And analyzing the congestion degree by calculating a ratio occupied by the foreground region in the road region from which the congestion analysis unit has been extracted. The method according to claim 1,
The step of extracting the background includes a step of extracting the background using the MOG2 algorithm after capturing the input frame and measuring the number of edges of the background to extract the background of the minimum edge. . The method according to claim 1,
Further comprising the step of the rotation detecting unit detecting the rotation of the CCTV,
A method for analyzing the road congestion by using the CCTV image which captures the next image to analyze the edge, performs the edge analysis again when the rotation is detected, and compares with the previous image and stores it in the corresponding image queue. In a road congestion analysis system robust against vehicle congestion using CCTV images,
A camera unit for photographing the road in real time;
A reference background extracting unit that extracts a background using a MOG2 algorithm after capturing an input frame of the captured image from the camera and measures a number of edges of the background to extract a background of a minimum edge;
A road region extracting unit for extracting a maximum edgeless area from the background extracted from the reference background extracting unit, extracting a road area after verifying the road area using the optical flow calculation result;
Extracting characteristics for each lattice by dividing the road area extracted from the road area extracting unit by a square lattice, extracting characteristics for each lattice of the current frame, calculating a difference in characteristics between the road area and the current frame, A foreground extracting unit for extracting foregrounds recognized as vehicles; And
And a congestion analyzer for analyzing the congestion degree by calculating a ratio of the foreground area in the extracted road area to the traffic congestion degree. 5. The method of claim 4,
And a rotation detecting unit for capturing the next image to perform edge analysis, performing edge analysis again when the rotation is detected, and comparing the previous image with the previous image and storing the same in the corresponding image queue. 6. The method of claim 5,
Wherein the rotation sensing unit is robust against vehicle congestion using a CCTV image that detects whether the vehicle is turning based on the occurrence frequency of the transverse edges. The method according to claim 1,
Wherein the step of analyzing the congestion level comprises: separating a trapezoidal area close to the camera from a trapezoidal area separated from the camera in a trapezoidal road area, wherein the width of the adjacent trapezoidal area is wider than the width of the trapezoidal area,
A method of analyzing a road congestion degree robust against a vehicle congestion using a CCTV image that estimates an area weight of a vehicle detected in a trapezoidal area spaced apart from a vehicle detected in the proximate trapezoidal area.

Images