KR20150125818A - Forward Vehicle Identification Method and System with the Vehicular Brake Lamp at Night Time - Google Patents

Abstract

Provided are a method for recognizing a forward vehicle at night by using a brake lamp of a vehicle and a system thereof. The method for recognizing a forward vehicle according to an embodiment of the present invention comprises the following steps of: extracting brake lamp areas of a vehicle in a forward image; and detecting a forward vehicle by coupling the extracted brake lamp areas. Therefore, vehicle detection strong for a night image is possible by detecting the forward vehicle in the night image by using the brake lamp of the vehicle in recognizing the forward vehicle and calculating a distance between vehicles by using the night image.

Description

TECHNICAL FIELD [0001] The present invention relates to a vehicle identification method and system using a vehicle brake,

The present invention relates to vehicle recognition, and more particularly, to a method and system for recognizing a vehicle ahead at night.

In order to recognize a forward vehicle using a camera installed in the vehicle, a feature of the forward vehicle must be extracted using a pre-collected vehicle DB, and a learning model should be trained using the extracted features.

Also, the forward vehicle candidate region extraction uses a trained learning device, and the extracted candidate regions pass the verification that can be regarded as a vehicle, and determine that the vehicle is the final vehicle.

Then, the distance between the own vehicle and the vehicle running ahead is calculated according to the determined size of the vehicle, and when the distance approaches, an alarm sound is generated to inform the driver of the collision risk information.

As shown in FIG. 1, the above description shows relatively excellent performance when the feature extraction of the vehicle is easy in the daytime image. However, in the case of nighttime, performance degradation can not be avoided when extracting the vehicle characteristics according to the above description.

As shown in Fig. 2, the background is dark at night, and feature extraction of the vehicle is not easy due to a blurring phenomenon such as a vehicle braking.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a nighttime ahead vehicle recognition method and system using a brake or the like of a vehicle in order to improve recognition performance of a front vehicle at night have.

According to an aspect of the present invention, there is provided a front vehicle recognizing method including extracting areas such as brakes of a vehicle from a forward image; And combining the extracted brake light areas to detect the front vehicle.

The extracting step may include generating a binary image by binarizing a forward image; Extracting an edge from a forward image to generate an edge image; And separating the candidate region by ANDing the binary image and the edge image.

The extracting may further include labeling the candidate regions to generate label regions; And separating the label regions in a part of the label regions, and the detecting step may combine the label regions separated in the separating step.

The sorting step may identify label regions of a specific size among the label regions.

Also, the detecting step may combine the areas of the extracted brakes in consideration of the symmetry, the size, the area, and the aspect ratio.

The detecting step may combine the extracted brake light areas in consideration of the outer color distribution of the extracted brake light area.

Further, the step of verifying the forward vehicle detected in the detecting step may be further performed by taking into consideration the motion information of the extracted brake and like areas.

According to another aspect of the present invention, there is provided a front vehicle recognition system comprising: a camera for generating a forward image; And an image processor for extracting braking areas of the vehicle from the forward image generated by the camera and combining the extracted braking areas to detect the forward vehicle.

As described above, according to the embodiments of the present invention, in the calculation of the headway distance by using the night image, the front vehicle is detected using the brake of the vehicle in the night image, Vehicle detection becomes possible.

Accordingly, the calculation of the headway distance is accurate even at night, and the safe operation of the driver who is negligent in forward sight can be achieved.

BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a drawing provided for a description of vehicle recognition performance in a daytime image,
2 is a diagram illustrating night vision perception performance,
3 is a flowchart provided in the description of the vehicle recognition and alarm method according to an embodiment of the present invention;
Fig. 4 is a view provided for explaining a candidate region detection process such as a vehicle braking,
Figures 5-7 illustrate the label area merging techniques,
FIG. 8 is a view provided for explaining a vehicle verification technique using motion information, and FIG.
9 is a block diagram of a front vehicle recognition system according to another embodiment of the present invention.

Hereinafter, the present invention will be described in detail with reference to the drawings.

1. Front vehicle recognition and alarm

3 is a flowchart provided in the description of the vehicle recognition and alarm method according to an embodiment of the present invention. The vehicle recognizing and alarming method according to an embodiment of the present invention accurately recognizes a vehicle from an image acquired through a camera at night and generates an alarm sound based on the calculated inter-vehicle distance based on the result.

As shown in FIG. 3, first, a forward image of the vehicle is acquired in real time through a camera installed in the vehicle (1).

Then, set the horizon according to the height of the vehicle and the installation height of the camera (②). This is to filter the vehicle candidate region appearing at a position higher than the horizon and enlarge the real-time image according to the obtained vehicle candidate region so as to recognize the vehicle at a long distance. Also, it is possible to reduce the front vehicle search area using the vanishing point information of the lane that can be obtained from the real time lane information.

Next, a preprocessing process of the input image is required in order to effectively extract the characteristics of the image necessary for vehicle recognition in the input forward image (3). Specifically, it improves the image below the horizon.

Histograms and nonlinear methods are used to improve the input image. For the nonlinear method, AINDANE (Adaptive and Integrated Neighborhood Dependent Approach for Nonlinear Enhancement) and HSV_AINDANE methods can be used.

Then, we extract the brake light area of the vehicle from the improved input image (④), and verify the actual vehicle by verifying the extracted brake light area (⑤). These steps are described in detail below.

Next, the size of the forward vehicle is determined, and the inter-vehicle distance is calculated through the LUT (LookUpTable) configured in advance (6). Then, the area-based tracking method is applied to track the vehicle ahead (⑦).

If it is determined that the distance from the preceding vehicle is close and the driver does not look ahead (8), a visual / audible alarm is generated to the driver (9). Whether or not the driver looks ahead is determined from the brake operation of the vehicle and the left and right direction indicators. If the driver is close to the vehicle and the driver is controlling the vehicle, it is not a dangerous situation. Even if the vehicle is controlled, the alarm sounds when the distance is very close to the predetermined speed.

2. Forward vehicle recognition at night

Hereinafter, a method for detecting the preceding vehicle will be described in detail through the steps " 4 "and" 5 "

2.1 Detection of candidate areas such as vehicle brakes

In order to extract the front vehicle area from the night image, the vehicle's brake light (taillight) is used. In the case of the night image, in order to solve the problem that the rear bumper area of the vehicle ahead is brightened due to the vehicle headlight of the vehicle, the bumper area and the area extracted by the light source are separated It is necessary.

To do this, image binarization using the OTSU technique is applied to the color space of the input image, and the thresholds applied to the binarization are different for each color space to extract the candidate regions such as the brakes. [A] in FIG. 4 is an input image, and [B] in FIG. 4 shows a binary image of an input image by the OSTU method.

Next, edges are extracted from the input image, and the candidate region is separated by ANDing the binary image and the edge image. [C] in Fig. 4 is the edge image of the input image, and [D] in Fig. 4 is the AND operation result of the binary image [B] and the edge image [C].

The AND operation of the binary image with the edge image is performed in order to separate the brake light and the vehicle bumper into one area because the vehicle headlight of the vehicle is reflected on the bumper of the vehicle ahead.

Thereafter, the separated candidate regions are labeled for each region, and the break areas are extracted through 'size filtering' for the labeled regions (label regions). 'Size filtering' excludes areas with very large or very small contrast sizes, such as brakes.

[E] in FIG. 4 shows the result of size filtering after labeling the [D] image.

2.2 Merging Label Areas

It is necessary to merge the areas (see [E] of Fig. 4) extracted as the brakes of the vehicle in order to consider them as one vehicle. In the case of the brake of a vehicle, it seems that the color of the red color is emphasized. However, in the actual image, the color of the red color is slightly higher than that of the near vehicle only.

Therefore, the regions after the brake light region extraction of the vehicle are merged based on the dictionary (experimental) rule. Specifically, the areas of the extracted brakes are divided into three areas: 1) symmetry (merging areas whose shapes are mutually symmetric), 2) similar size / area (merging regions of similar size or area), 3) 4) similar illumination information (merging similar areas of illumination information), 5) size according to distance (combining regions with similar size according to distance), 6) correlation outside the region The merging is performed.

FIG. 5 shows a label area merging technique based on symmetry, similar size (area), similar aspect ratio, similar illumination information, and FIG. 6 shows a label area merging technique based on a variable size according to the distance FIG. 7 shows a label area merging technique based on similar correlation outside the area.

Similar correlation outside the area means that the color distribution outside the label area extracted by the brakes is red. That is, it merges the label areas whose outer color distribution is red.

2.3 Vehicle Verification

The label area tracking method is used as a method for verifying the merged one vehicle. The label of the area merged with one vehicle uses the feature that the same motion information appears.

In the case of a complicated city center, the movement of each lighting is different due to the fixed ambient lighting and the movement of the moving forward vehicles. In the case of a label merged into one vehicle, if the same motion information is generated during N frames, it is verified with an actual vehicle.

FIG. 8 shows a vehicle verification technique using motion information.

3. Forward vehicle recognition system

9 is a block diagram of a front vehicle recognition system according to another embodiment of the present invention. 9, the front vehicle recognition system includes a camera 110, a display 120, an image processing unit 130, a storage unit 140, a control unit 150, and an output unit 160. As shown in FIG.

The camera 110 acquires the forward image of the vehicle in real time, and the display 120 displays the captured forward image and other information. The storage unit 140 stores a forward image acquired by the camera 110. [

The image processor 130 performs horizon setting, image preprocessing, area extraction such as braking of the front vehicle, vehicle verification, etc., determines the size of the front vehicle, calculates the inter-vehicle distance, and tracks the front vehicle.

The extraction / calculation / tracking result of the image processing unit 130 may be displayed overlapping with the forward image on the display 120.

The control unit 150 generates a visual / audible alarm to the driver through the output unit 160 when it is determined that the distance to the preceding vehicle is close and the driver is not looking ahead.

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, but, on the contrary, It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the present invention.

110: camera
130:
150:
160: Output section

Claims (8)

Extracting braking areas of the vehicle from the forward image; And
And combining the extracted areas of the brake and detecting the front vehicle.
The method according to claim 1,
Wherein the extracting step comprises:
Generating a binary image by binarizing a forward image;
Extracting an edge from a forward image to generate an edge image; And
And performing an AND operation on the binary image and the edge image to separate the candidate region.
3. The method of claim 2,
Wherein the extracting step comprises:
Labeling the candidate regions to generate label regions; And
Further comprising the steps of: identifying label regions of a portion of the label regions,
Wherein the detecting step comprises:
And combining the label regions separated in the classification step.
The method of claim 3,
Wherein,
And label regions belonging to a specific size range of the label regions are distinguished.
The method according to claim 1,
Wherein the detecting step comprises:
Wherein the extracted areas of the brake are combined in consideration of symmetry, size, area, and aspect ratio.
The method according to claim 1,
Wherein the detecting step comprises:
And the extracted brake light areas are combined in consideration of the outer color distribution of the extracted brake light area.
The method according to claim 1,
And verifying the forward vehicle detected in the detecting step, taking into consideration the motion information of the extracted areas of the brake and the like.
A camera for generating a forward image; And
And an image processor for extracting braking areas of the vehicle from the forward image generated by the camera and combining the extracted braking areas to detect a forward vehicle.

Images