KR20160035460A - System for extracting vehicle candidate group and method thereof - Google Patents

System for extracting vehicle candidate group and method thereof Download PDF

Info

Publication number
KR20160035460A
KR20160035460A KR1020140127100A KR20140127100A KR20160035460A KR 20160035460 A KR20160035460 A KR 20160035460A KR 1020140127100 A KR1020140127100 A KR 1020140127100A KR 20140127100 A KR20140127100 A KR 20140127100A KR 20160035460 A KR20160035460 A KR 20160035460A
Authority
KR
South Korea
Prior art keywords
vehicle
candidate
horizontal edge
unit
clustering
Prior art date
Application number
KR1020140127100A
Other languages
Korean (ko)
Inventor
최진혁
박태곤
권구도
오동언
신동인
Original Assignee
현대자동차주식회사
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 현대자동차주식회사 filed Critical 현대자동차주식회사
Priority to KR1020140127100A priority Critical patent/KR20160035460A/en
Publication of KR20160035460A publication Critical patent/KR20160035460A/en

Links

Images

Classifications

    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/052—Detecting movement of traffic to be counted or controlled with provision for determining speed or overspeed

Abstract

The present invention relates to a system and a method thereof to detect a vehicle candidate group at high speed and, more specifically, to technology, capable of detecting a surrounding vehicle candidate group of a vehicle of a user at high speed by using a horizontal edge component and camera geometry information. According to an embodiment of the present invention, the system includes: an image obtaining part obtaining a surrounding image of a vehicle; and a vehicle candidate group detecting device extracting the horizontal edge component from the image data, received from the image obtaining part, detecting a candidate line by using the continuity of the horizontal edge component, performing clustering by candidate line, and then performing filtering to detect a surrounding vehicle candidate group.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention [0001] The present invention relates to a high-

The present invention relates to a high-speed vehicle detection system and a method thereof, and more particularly, to a technology for detecting a vehicle candidate group around a vehicle at high speed using camera geometry information and a horizontal edge component.

With the development of vehicle technology, vehicle safe driving systems for safe driving such as distance detection between vehicles and collision detection between vehicles have been continuously developed.

In order to increase the reliability of such vehicle safe driving systems, it is important to accurately detect vehicle candidates around the vehicle.

Conventionally, 2D or 3D position information of a vehicle candidate group around the vehicle is extracted using a radar, a radar, an ultrasonic wave, a camera, and the like.

However, since the radar and the radar are costly equipment, the cost burden is large and the ultrasonic wave is limited to the distance, and the algorithm using the image information using the camera has a complicated algorithm.

That is, the horizontal and vertical edge components are detected in the image information and the largely intersected portion is determined as the vehicle position on the floor. In order to identify the intersection, a peak is searched through a histogram of edges, A complicated algorithm such as an application step, a step of verifying a left / right pointer that is suspected to be a vehicle is required.

 In particular, since the resolution of a camera is gradually increasing, when the above-described algorithm is applied to an entire area having a high resolution, much time and cost are consumed.

An embodiment of the present invention is to provide a vehicle candidate group high-speed detection system and a method thereof capable of detecting a vehicle candidate group at a high speed.

The vehicle candidate group high speed detection system according to an embodiment of the present invention includes an image acquisition unit for acquiring an image around a vehicle; And extracting a horizontal edge component from the image data received from the image obtaining unit, detecting candidate lines by using the continuity of the horizontal edge components, performing clustering on the candidate lines, filtering the neighboring vehicle candidates, And may include a vehicle candidate group detection device for detecting the vehicle candidate group.

The vehicle candidate group detection device may further include: an image data collection unit that receives image data from the image acquisition unit and sets an area of interest in the image data; A horizontal edge component extraction unit for extracting a horizontal edge component from the ROI and removing the noise using continuity among the horizontal edge components to detect the candidate line; A clustering unit for performing clustering for each candidate line; A filtering unit for performing filtering on the clustered clusters using a filtering table for a width of a cluster for each Y coordinate according to a camera geometry; And a vehicle candidate group detection unit for detecting the filtered clusters as a vehicle candidate group.

The clustering unit may perform clustering for each candidate line to detect clusters, and perform clustering again between the detected clusters.

In addition, the filtering table may be information obtained by matching cluster widths according to Y coordinates, and may further include a storage unit for storing the filtering table.

The image acquiring unit may be a camera.

According to another aspect of the present invention, there is provided a method for detecting a vehicle candidate group, the method comprising: calculating a horizontal edge component from image data of a vehicle; Detecting a candidate line from the horizontal edge component; Performing clustering for each candidate line to detect clusters; And performing filtering on the detected clusters.

The step of detecting the candidate line may include determining that the continuity of the horizontal edge components is equal to or greater than a predetermined reference value, and determining that the horizontal edge components are noise if the continuity is less than a predetermined reference value.

The filtering may include defining a reference value for the width of the cluster according to the Y coordinate of the image data, and if the width of the detected clusters satisfies the reference value, And judges that the detected clusters are not vehicles when the width of the detected clusters does not satisfy the reference value.

This technology can be applied to various vehicle safety systems using vehicle candidate detection while reducing cost by enabling high-speed detection of vehicle candidates with only camera sensor.

1 is a configuration diagram of a vehicle candidate group high-speed detection system according to an embodiment of the present invention.
FIG. 2 is a flowchart showing a method of detecting a vehicle candidate group at a high speed according to an embodiment of the present invention.
FIG. 3 is an example of setting a region of interest using camera geometry in image data.
4 is an example of extracting a horizontal edge component from image data.
5 is an exemplary diagram illustrating noise removal and continuity lines extracted from image data obtained by extracting horizontal edge components.
FIG. 6 is an example of clustering by perimeter lines in image data.
FIG. 7 is a diagram illustrating clustering and cluster filtering performed by clustering per peripheral line in FIG. 6; FIG.
8A and 8B are diagrams for explaining the cluster filtering.
FIG. 9 is a diagram illustrating an example in which a bounding box process is performed on a candidate candidate group after cluster filtering. FIG.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings, in order to facilitate a person skilled in the art to easily carry out the technical idea of the present invention.

The present invention discloses a technique capable of detecting a candidate vehicle group at high speed using only a camera.

Hereinafter, embodiments of the present invention will be described in detail with reference to FIGS. 1 to 9. FIG.

1 is a configuration diagram of a vehicle candidate group high-speed detection system according to an embodiment of the present invention.

The vehicle candidate group high speed detection system according to an embodiment of the present invention includes an image acquisition unit 100, a vehicle candidate group detection device 200, a display unit 300, and a storage unit 400.

The image obtaining unit 100 photographs the surrounding environment such as the front, rear, and side of the vehicle. For this, the image acquisition unit 100 may include a camera or the like. At this time, the image acquiring unit 100 does not need to separately provide a camera installed in a vehicle safety system such as a conventional lane departure warning system (LDWS).

The vehicle candidate group detection apparatus 200 obtains horizontal edge components from the image data received by the image obtaining unit 100, detects candidate lines, performs line-by-line clustering, performs clustering between clusters again, performs cluster filtering And performs the bounding box process to detect the candidate vehicle group at a high speed.

The vehicle candidate group detection apparatus 200 includes an image data collection unit 210, a horizontal edge component extraction unit 220, a clustering unit 230, a filtering unit 240, and a vehicle candidate group detection unit 250.

The image data collection unit 210 receives the image data from the image acquisition unit 100 and sets an ROI for the camera geometry with respect to the image data as shown in FIG.

The horizontal edge component extraction unit 220 calculates a horizontal edge component using a sobel mask for a region of interest set in the image data collection unit 210 as shown in FIG. At this time, although the example of calculating the horizontal edge component using the bevel mask is described in the present invention, the horizontal edge component can be calculated by applying various methods for calculating the horizontal edge component.

At this time, the bevel mask is expressed by the following equations (1) to (3). As shown in Equation (1), the X coordinate and the Y coordinate are separated to calculate the size (Gx) for the X axis and the size (Gy) for the Y axis.

Figure pat00001

In Equation (1), Gx is the size with respect to the X axis, Gx is the size with respect to the Y axis, and A is the portion corresponding to the size of 3 by 3 of the image data here. That is, in the case of image data having a resolution of 640 * 480, 640 * 480 size is divided by 3 by 3 and substituted into A value.

Figure pat00002

G, which is the magnitude (intensity) of the edge, can be calculated by using Equation 2 using Gx and Gy calculated in Equation (1).

Figure pat00003

Further, using the Gx and Gy calculated in the equation (1), the direction of the edge,?, Can be calculated as shown in equation (3).

In addition, when the horizontal edge is detected as shown in FIG. 4, the horizontal edge component extraction unit 220 extracts the continuity line by removing the noise using the continuity of the horizontal edge as shown in FIG. That is, the horizontal edge component extraction unit 220 determines that the horizontal edge component is noise if the line is less than a predetermined length, and extracts it as a continuous line when the length is equal to or longer than a predetermined length. In Fig. 5, the white line portion corresponds to the continuity line.

The clustering unit 230 performs clustering on the continuity lines of the image data to detect neighboring clusters. In this case, the nearest neighbors clustering technique is used. The closest clustering technique is a general technique, so a detailed description will be omitted. At this time, when the image data is 640 * 480 resolution as shown in FIG. 4, clustering is sequentially performed on (0, 0) coordinates to (640, 480) coordinates.

In addition, the clustering unit 230 performs clustering between the detected neighboring clusters. At this time, clustering between clusters can also be performed using closest clustering method.

The filtering unit 240 performs cluster filtering using camera geometry in the image data clustered between the clustering units.

That is, the filtering table is stored as shown in Table 1 below.

Y coordinate Width Threshold 300 One One ... ... ... 400 138 24 ... ... ... 570 450 100

At this time, the filtering table includes information about the width of the vehicle according to the Y coordinate in the image data and a threshold value including an error range therefor. That is, considering the distance between the vehicle and the surrounding vehicle, the width of the vehicle ahead may be narrower as the distance increases. Actual image data has X coordinate and Y coordinate, and the width of the vehicle according to Y coordinate is stored in advance as an experimental value. If the detected line width with respect to the Y coordinate is too wide or too narrow, it can be judged to be not a vehicle.

The vehicle candidate group detection unit 250 detects a portion detected as a line as a vehicle candidate group. At this time, the vehicle candidate group detection unit 250 may perform a rounding box process on the detected vehicle candidate group so as to be displayed on the screen.

The display unit 300 can display a video image on which a vehicle candidate group is detected or a vehicle safety service related screen through detection of a vehicle candidate group.

The storage unit 400 stores the image data collected by the image obtaining unit 100 for a predetermined time and stores a filtering table that matches the width and the threshold according to Y coordinates for filtering as shown in Table 1. [

Hereinafter, a vehicle candidate group high speed detection method according to an embodiment of the present invention will be described in detail with reference to FIG.

First, the image acquisition unit 100 transmits the image data photographed around the subject to the image data collection unit 210 (S101). The image data collection unit 210 sets an ROI for the camera geometry as shown in FIG. 3 (S102). At this time, a general technique is used to set the region of interest for the camera geometry.

Thereafter, the horizontal edge component extraction unit 220 extracts horizontal edge components on the ROI (S103). At this time, the horizontal edge component is extracted using a bevel mask as shown in the above-mentioned Equations (1) to (3). In this case, an example in which a horizontal edge component is extracted is shown in FIG.

Next, the horizontal edge component extraction unit 220 checks the continuity of the X axis among the extracted horizontal edge components, removes the noise, and extracts the continuity line (S104). In other words, the horizontal edge component extractor 220 determines that noise is continuous if the value of the horizontal edge component continuous to the X axis is less than a predetermined reference value and removes the noise. If the continuous value in the X axis is greater than or equal to a predetermined reference value, Line. In this case, FIG. 3 shows an example in which the noise is removed and only the continuity line is extracted.

Thereafter, the clustering unit 230 performs clustering for each line (S105). At this time, nearest neighbors clustering is used for clustering, and a recent clustering technique is a general clustering technique, so a detailed description thereof will be omitted. Here, FIG. 6 shows an example of clustering per peripheral line.

Then, the clustering unit 230 again performs clustering between clusters in step S105 (S106). Clustering at this time also uses the recent room clustering technique. FIG. 7 shows an example in which clustering is performed between clusters.

Thereafter, the filtering unit 240 performs clustering filtering using the camera geometry on the clustering-processed image data (S107). At this time, the filtering may be performed using the filtering table of Table 1 described above.

The filtering table is a table in which a width (width) of a horizontal edge cluster according to the Y coordinate of the camera image is predefined. For example, when the Y coordinate is 300, the vehicle is recognized when the width of the cluster is 1. When the Y coordinate is 400, the vehicle is recognized when the detected width of the cluster is 138. The relationship graph of the Y coordinate and the cluster width is as shown in FIG. 8B, and an example of the relationship is shown in FIG. 8A. However, in order to consider the error when determining the width of the cluster for each Y coordinate, the threshold values are stored together in Table 1. That is, when the Y coordinate is 400, since the width of the cluster is 138 and the threshold is 24, the vehicle is recognized when the width of the cluster is in the range of 114 to 162. FIG. 8B is a graph considering such a threshold value. Referring to FIGS. 8A and 8B, it can be seen that as the Y value increases, the width of the vehicle candidate zone increases gradually.

Through this filtering, it is possible to detect more accurate vehicle candidate group by judging whether or not the detected cluster is a vehicle. In this case, FIG. 7 shows an example of filtering after clustering between clusters. That is, a total of 21 clusters are searched in FIG. 6, and it is understood that after filtering is performed in FIG. 7, it is reduced to 15 clusters.

Subsequently, the vehicle candidate group detection unit judges the finally detected clusters as the final vehicle candidate group (S108).

As described above, according to the present invention, it is possible to accurately detect a vehicle candidate group using only a camera sensor without using a conventional ultrasonic sensor, a radar, a radar, etc., The functions of the present invention may be mounted and used.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, It should be regarded as belonging to the claims.

Claims (8)

An image acquiring unit for acquiring images of the surroundings of the vehicle; And
A horizontal edge component extraction unit for extracting a horizontal edge component from the image data received from the image obtaining unit, detecting candidate lines by using the continuity of the horizontal edge components, performing clustering for each of the candidate lines, Vehicle candidate detection unit
The high-speed detection system comprising: The method according to claim 1,
The vehicle candidate group detection apparatus (100)
An image data collecting unit receiving image data from the image acquiring unit and setting an area of interest in the image data;
A horizontal edge component extraction unit for extracting a horizontal edge component from the ROI and removing the noise using continuity among the horizontal edge components to detect the candidate line;
A clustering unit for performing clustering for each candidate line;
A filtering unit for performing filtering on the clustered clusters using a filtering table for a width of a cluster for each Y coordinate according to a camera geometry; And
A vehicle candidate group detection unit for detecting the filtered clusters as a vehicle candidate group,
Further comprising: means for determining whether or not the vehicle is a candidate vehicle. The method of claim 2,
The clustering unit,
Detecting clusters by performing the candidate line-by-line clustering, and clustering the detected clusters again. The method of claim 2,
The filtering table is information obtained by matching cluster widths according to Y coordinates,
And a storage unit for storing the filtering table. The method according to claim 1,
The image acquiring unit may acquire,
And the camera is a camera. Calculating a horizontal edge component from image data captured around the vehicle;
Detecting a candidate line from the horizontal edge component;
Performing clustering for each candidate line to detect clusters; And
Performing filtering on the detected clusters
And determining whether the candidate candidate group is a candidate candidate group. The method of claim 6,
Wherein the step of detecting the candidate line comprises:
Wherein when the continuity among the horizontal edge components is greater than or equal to a predetermined reference value, the vehicle candidate is detected, and if the continuity among the horizontal edge components is less than a predetermined reference value, noise is determined to be eliminated. The method of claim 6,
The step of performing the filtering includes:
A reference value for the width of the cluster according to the Y coordinate of the image data is defined in advance; if the width of the detected clusters satisfies the reference value, it is determined that the detected clusters are vehicles; And judges that the detected clusters are not vehicles if the reference value is not satisfied.
KR1020140127100A 2014-09-23 2014-09-23 System for extracting vehicle candidate group and method thereof KR20160035460A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
KR1020140127100A KR20160035460A (en) 2014-09-23 2014-09-23 System for extracting vehicle candidate group and method thereof

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
KR1020140127100A KR20160035460A (en) 2014-09-23 2014-09-23 System for extracting vehicle candidate group and method thereof

Publications (1)

Publication Number Publication Date
KR20160035460A true KR20160035460A (en) 2016-03-31

Family

ID=55652163

Family Applications (1)

Application Number Title Priority Date Filing Date
KR1020140127100A KR20160035460A (en) 2014-09-23 2014-09-23 System for extracting vehicle candidate group and method thereof

Country Status (1)

Country Link
KR (1) KR20160035460A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2019212113A1 (en) * 2018-05-03 2019-11-07 충북대학교 산학협력단 Density-based object detection device and method using lidar sensor

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2019212113A1 (en) * 2018-05-03 2019-11-07 충북대학교 산학협력단 Density-based object detection device and method using lidar sensor

Similar Documents

Publication Publication Date Title
JP4930046B2 (en) Road surface discrimination method and road surface discrimination device
CN107750213B (en) Front vehicle collision warning device and warning method
US10339396B2 (en) Vehicle accessibility determination device
JP2006268097A (en) On-vehicle object detecting device, and object detecting method
JP7206583B2 (en) Information processing device, imaging device, device control system, moving object, information processing method and program
JP5959073B2 (en) Detection device, detection method, and program
JP4872769B2 (en) Road surface discrimination device and road surface discrimination method
US11691585B2 (en) Image processing apparatus, imaging device, moving body device control system, image processing method, and program product
JP2008165765A (en) Method and apparatus for acquiring vehicle side image, vehicle lamp recognition error detection method and prediction method for safe driving
EP3163506A1 (en) Method for stereo map generation with novel optical resolutions
WO2014002692A1 (en) Stereo camera
JP6592991B2 (en) Object detection apparatus, object detection method, and program
JPH11351862A (en) Foregoing vehicle detecting method and equipment
CN109886064B (en) Method for determining the boundary of a drivable space
JP2000207693A (en) Obstacle detector on vehicle
EP2677462B1 (en) Method and apparatus for segmenting object area
JP2007272292A (en) Shadow recognition method and shadow boundary extraction method
JP7180445B2 (en) object detector
Romdhane et al. A generic obstacle detection method for collision avoidance
JP2008004989A (en) Vehicle perimeter display apparatus
EP3410345A1 (en) Information processing apparatus and non-transitory recording medium storing thereon a computer program
KR101205565B1 (en) Method for Dectecting Front and Rear Vehicle by Using Image
JP2008165595A (en) Obstacle detection method, obstacle detection device, and obstacle detection system
KR101501531B1 (en) Stereo Vision-based Pedestrian Detection System and the method of
KR102188164B1 (en) Method of Road Recognition using 3D Data

Legal Events

Date Code Title Description
A201 Request for examination
E902 Notification of reason for refusal
E601 Decision to refuse application